Jeffrey Jay Rutgard appeals his conviction of numerous counts of mail fraud on, and false claims to, Medicare, of counts of mail fraud on other insurers, and of transactions in money derived from the frauds. He appeals as well a judgment of forfeiture. He also appeals his sentence, challenging the enhancements arid the calculation of loss that led to the imposition of a sentence of 11% years and an order of restitution of over $16 million embracing virtually the entire proceeds of his practice as an ophthalmologic surgeon between 1988 and April, 1992.

This case, hard fought at trial and on appeal, with a vigorous attack on the sufficiency of evidence supporting conviction and presenting a transeript based on five months of trial, has required this court to determine to what degree a defense of good faith belief in the necessity of medical treatment may be based on disputed expert testimony. It has required this court to decide whether entries made by a physician in his patient’s files fall within federal jurisdiction. It has required this court to determine whether proof of particular examples of fraud in billing insurers establishes that the practice is, as a whole, a scheme to defraud. It has required us to decide whether 18 U.S.C. § 1957 , the statute forbidding certain bank transfers, is violated if the transfers involve commingled criminal and innocent funds. Holding that a reasonable doubt may be created as to good faith when qualified experts disagree as to medical necessity, that statements in the files are within the jurisdiction of the government when the government as insurer must rely on them to document medical necessity, that fraud in specific cases and in specific kinds of billing does not prove that the whole practice is a fraud, and that proof of violation of § 1957 requires proof of the transfer of particular criminal proceeds, we affirm certain counts of conviction, reverse others, set aside the order of forfeiture, vacate the sentence, and remand for resentencing.

BACKGROUND

The Insurers. Medicare is the federal health insurance program for persons over the age of 65 who meet its qualifications. 42 U.S.C. § 1395o. As Medicare beneficiaries are 65 years of age or older, medical treatment of their eyes is a significant part of Medicare coverage. For example, in 1991, a year involved in this case, it is estimated that 45 percent of the approximately thirty million beneficiaries of Medicare received eye care paid for by Medicare, which processed thirty-five million claims for such care. Leon B. Ellwein, et al., Use of Eye Care Services Among The 1991 Medicare Population, 103 Ophthalmology 1732, 1734 (1996). Cataracts were the most common eye condition, accounting for 41 percent of the patients’ visits and 1.2 million cases of surgery. Id. at 1732. The cost was over 12 percent of Medicare’s total budget. Id. In 1991 cataract surgery was the largest single outpatient category funded by Medicare. See American Academy of Ophthalmology v. Sullivan, 998 F.2d 377 , 379 (6th Cir.1993). The cost demonstrated by these statistics underlines the importance to the government and to taxpayers of honest billing in the practice of ophthalmologists, and they illustrate the prevalence of conditions among those over the age of 65 that necessitate treatment of their eyes.

Medicare pays only for services that are “reasonable and necessary for the diagnosis or treatment of illness or injury or to improve the functioning of a malformed body member.” 42 U.S.C. § 1395y(a)(l)(A). Participating providers are required to ensure that any services rendered to Medicare recipients are supported by sufficient evidence of medical necessity. 42 U.S.C. § 1320c-5(a)(1). Under the statute, the Secretary of Health and Human Services regulates the administration of the program with regulations now totalling over 1,600 pages. The agent of Medicare in Southern California is Transamerica Occidental Life Insurance Company.

CHAMPUS is the federal health insurance program for Department of Defense employees. Like Medicare, CHAMPUS pays only for medically necessary services and supplies. 32 C.F.R. § 199.4 (a)(1)®. The CHAMPUS insurance carrier is Foundation Health.

The Railroad Retirement Health Insurance Program is the federal health insurance program for retired railroad workers administered by the Railroad Retirement Board. Travelers Insurance Company serves as the insurance carrier. The Railroad Retirement plan follows the same rules and regulations as Medicare.

Blue Cross of Southern California and Aet-na are insurers not connected with the government.

Jeffrey Jay Rutgard. Dr. Rutgard, the defendant, was born in Chicago, Illinois in 1951. His father was a physician. He attended the College of Medicine of the University of Illinois. He also enrolled in the university’s graduate school where he was a James Scholar. On the basis of his proficiency in these schools he skipped internship and moved directly to a three-year residency in ophthalmology at the Medical School of the University of Iowa. In 1981 he came from this residency to practice in San Diego, California.

In 1982 he secured his own medical practice by the purchase of Dr. Amos Root’s in San Diego. Four years later he bought Dr. John Bickerton’s in La Jolla and thereafter had two offices. In 1989 he opened as well an ambulatory surgery center in La Jolla. His practice flourished. He kept an extraordinary schedule of working 6/> days per week, 16 hours a day, with trips to medical meetings but without vacations. He also made a great deal of money. From 1988 through May, 1992 he received from Medicare (80% of his practice) over $15 1/2 million.

His skill as a surgeon was recognized. The success of his practice he attributed in large measure to word-of-mouth recommendation by satisfied patients. Dr. Bickerton and several other ophthalmologists sent members of their families to him for surgery. Members of the staff who were government witnesses at trial had only praise for his surgical abilities.

Rutgard was also an active promoter of his practice. One-third of his practice he estimated came from his “community relations” program. He employed staff to hold free screening at senior citizen centers and nursing and retirement homes and, although without professional qualifications, to diagnose cataract and eyelid problems, suggesting Rutgard as the doctor who could cure them. Vans arranged by Rutgard brought this kind of clientele to his climes for further examination. Members of his staff, including the van drivers and the receptionist, were paid bonuses or qualified for raises by the number of the persons they successfully recruited to have surgery by him. In pursuit of these rewards some staff would tell strangers wearing glasses whom they saw in a supermarket that they should see Dr. Rut-gard.

Rutgard approved his office manager, George Butera, setting “quantifiable goals” for the practice. For example, for 1991, the quantifiable goals were as follows:

The quantifiable goals for 1992 were increased in each category. For example, the goal for YAGs was increased to 1,175 and for ptosis surgeries to 960.

Rutgard was not shy about self-promotion. Staff was instructed to say, “Doctor is the best eye doctor in San Diego.” Staff was told to tell patients, “The Lord guides his work and skills.” Rutgard was rebuked by the American College of Eye Surgeons for putting out a resume stating that he was a fellow rather than a member of that college. He was criticized in 1991 by the chairman of the ethics committee of the American Academy of Ophthalmology for an advertisement stating that he was “board-certified” by this academy, which does not do such certification. He gave patients Cataracts And Implants. A Guide For Patients And Their Families. The front, inside and back covers all showed him as the co-author of the book with Dr. Phillip C. Hoopes, and the book was dedicated by the two authors to their wives. In fact only Dr. Hoopes was the author, and Rutgard’s name appeared as author because he paid Hoopes to have his name displayed as author.

PROCEEDINGS

How Rutgard became the subject of a federal investigation is not part of the record in this case. On March 24,1994 a 217-count indictment against Rutgard was returned by a federal grand jury. The district court scheduled trial for June. On May 4, the court declared the case complex and reset the trial for August. On June 8, the court re-set the trial date for October 4 and admonished counsel that the date was firm.

At the arraignment Rutgard was represented by Richard “Racehorse” Haynes, a member of the Texas bar admitted pro hac vice to practice in California. Haynes had also recently represented Rutgard in lengthy license revocation proceedings involving many of the issues covered in the indictment. Haynes was joined by local counsel Juanita Brooks, an active member of the San Diego bar.

The first major procedural battle occurred on September 19 when Rutgard pro per moved to relieve Haynes and to secure a continuance. In support of the motion Haynes filed a declaration that he had not read the “thousands of pages” of documents supplied since August 19 by the government and could not physically do so by October 4 and that in consequence he would be unprepared to render effective assistance as counsel on the scheduled trial date. Brooks declared that she was only prepared to “assist ” Haynes.

The district court denied the motion to relieve Haynes and to continue the trial date. The court noted that Haynes had been involved with the- case “since before its inception” and was aware of its complex character. The court noted that as early as May 20, 1994 Brooks was aware of Haynes’ lack of diligence in preparing. Haynes was unprepared now because he had “simply chosen to spend his time doing other things.” There was still two weeks’ time for Brooks and her firm to prepare the case. The inconvenience to the court in resetting a trial expected to last at least two months was serious, as was the inconvenience to the government.

With the continuance denied, Haynes went on as lead counsel. On October 3 he developed a painful abscessed tooth and moved for a 60-day continuance. The court denied the motion. Thereupon Brooks, although declaring herself unprepared, participated in the selection of the jury. When selection was completed, the case was put over to October 12 to permit Haynes to return from an operation on his tooth.

At voir dire of the prospective jurors Rut-gard moved to have the court make specific inquiry as to whether the jurors had been exposed to hostile coverage of his problems in the media. He stated that there had been at least six articles in the Los Angeles Times, ten in the San Diego Union Tribune, and fourteen in the LaJolla Light that had been run and created unfavorable publicity. The court merely inquired if any member of the jury panel had “read or heard anything about this case” other than the indictment. There was no affirmative response.

The court ruled in limine that four ex-employees of Rutgard testifying for the government could not be cross-examined on the civil suit they had pending against him seeking compensatory and punitive damages for malicious prosecution unless the government then had the opportunity to bring out the circumstances of the suit. The background of the suit was that these employees had testified against Rutgard in the state proceedings for revocation of his license, which had resulted in the revocation; and that he sued the employees for slander in the federal district court and the suit was dismissed in February, 1994. Faced with this ruling, the defense chose not to examine the employees on their suit. Later in the trial the defense suggested that an employee or someone connected with an employee might have filed a qui tarn action against Rutgard. The government filed a response under seal.

The trial proceeded for the next five months. The court made a number of comments and rulings to which the defense took exception and which now form part of the grounds for appeal. In particular, when a government witness was explaining what she meant by her use of “honesty and integrity,” Haynes made an evidentiary objection. The court said, “You see the word ‘honesty’?” Haynes: ‘Tes, I do.” The court: “All right. That’s what she is explaining. It’s called honesty and integrity.” Haynes: “I understand, Your Honor.” The court: “I hope so. Objection is overruled.” Haynes sought a sidebar, objecting that the court’s remark would be understood by the jury as a shot at his client. Overruling another objection to evidence by Haynes, the court said: ‘Tour objections are noted for the record. Take it up with the Ninth Circuit.” This comment was made in the presence of the jury.

After a month of trial, the court was dissatisfied with the pace at which the trial was moving. The court said to counsel outside the presence of the jury: “[T]here are going to be ultimatums laid down to both sides, time ultimatums .... [a]nd I’m going to enforce that and cut you off, and I don’t care if the Ninth Circuit likes it or not, and I don’t care whether the Supreme Court likes it or not.” The time limits were three hours apiece on direct and cross-examination and one hour apiece on redirect and recross. Violation would be a contempt of court. These time limits were thereafter in force.

During trial the government dismissed counts involving six patients, thereby reducing the number of counts to 132. On March 15, 1995 the jury found Rutgard guilty on each remaining count. The jury then found the amounts of $5,629,220.74 and $1,935,-220.48 to be moneys derived from the frauds and transferred by wire in violation of 18 U.S.C. § 1957 and found the amounts forfei-table under 18 U.S.C. § 982 (a)(1). On June 26, 1995 the court imposed a sentence with enhancements for vulnerable victims, abuse of trust, leadership role, perjury on the witness stand, and loss determined by Rutgard’s receipts from 1988 to April, 1992 when the government effectively closed his office.

Rutgard appeals.

ANALYSIS

RULINGS OF THE TRIAL COURT

Rutgard argues that cumulatively the errors of the district court deprived him of a fair trial and explain why the jury convicted him despite the insufficiency of the evidence. Reviewing the record, we find no single error or combination of errors that would sustain Rutgard’s contentions.

The ruling on the request to relieve Haynes and to continue the trial date was a ruling within the discretion of the district court, requiring the careful examination and balancing of various considerations which the court explicitly made, See United States v. Robinson, 967 F.2d 287, 291 (9th Cir.1992).

With his motion to examine the jurors specifically as to their exposure to hostile media coverage, Rutgard gave the number of articles that had been published, but did not attach the articles. On this appeal he asks us to take judicial notice of the articles as we did in Dockray v. Phelps Dodge Corp., 801 F.2d 1149 , 1152 n. 3 (9th Cir.1986). We have no obligation to take such notice when the contents were not brought before the district court. However, in the interest of fairness we have examined the articles attached to Rutgard’s motion requesting judicial notice.

The newspaper articles fall into two classes: those written before trial and those written during trial. Those written during trial recapitulate court proceedings and testimony the jury itself heard. While occasionally the articles might give an emphasis beyond what a juror would have noted in the courtroom, the net impact is de minimis and could not have prejudiced a juror during trial. The pretrial articles do contain inflammatory1 material of a prejudicial character, largely, it would appear, coming from the state Medical Board or its representatives, and once, it would appear, coming from the United - States Attorney’s office. If a juror had read this material, it is fair to say that he or she would have had a highly unfavorable view of Rutgard as a cheat. The coverage, however, was. not extensive. It was intermittent, triggered by. various events such as the federal raid in April, ,1992 on Rutgard’s home and office and by the state proceedings for suspension of his license in May, 1992 and the state’s actual revocation in May, 1994. Except for one front page story in the San Diego Union Tnbune about how several doctors, including Rutgard, were receiving disability payments for stress after being forced from practice by charges against them, the stories were on the inner pages of the metropolitan papers. The headlines typically referred to “a doctor” or “surgeon,” although several of the stories carried photos of Rutgard with his name and the stories uniformly identified him by the second paragraph. Unless a juror were particularly interested in medical fraud, it is not likely that this sporadic coverage would have generated any memory.

The defense could have made a stronger showing of potential prejudice by producing the articles before the district court and the district court could have made a more particular inquiry of potential jurors than in fact it did. The question, however, is whether the district court abused its discretion in not going beyond the usual inquiry as to whether the jurors had read anything about the case. United States v. Baker, 10 F.3d 1374, 1403 (9th Cir.1993); United States v. Giese, 597 F.2d 1170, 1183-84 (9th Cir.1979). We conclude that it did not.

The district court’s ruling on the cross-examination of the four ex-employees, who were described by Rutgard as “key” government witnesses, struck a fair balance. To have permitted cross-examination to develop a hostile motive to testify against Rutgard would have given only half the picture when the original hostility began with Rutgard’s resentment of their testimony against him in the license revocation proceeding. The court was entitled to use its discretion in ruling on how much of the relevant circumstances could be brought out by the government if it had to rehabilitate its witnesses. See United States v. Shapiro, 879 F.2d 468, 472-73 (9th Cir.1989).

As to the possibility of qui tam the government filed its response under seal because it believed that written disclosure of the existence of any such actions would violate the secrecy that protects an informant. At the request of Rutgard on this appeal we have inspected the sealed response of the government and are satisfied that as of the time of the response, January 17, 1995, there were no qui tam actions filed by anyone seeking a reward for information provided against Rut-gard.

In a long trial, marked by repetitive cross-examination by the defense, the court had the discretion to impose reasonable time limits on the examination of the witnesses. See Wood v. Alaska, 957 F.2d 1544, 1549-50 (9th Cir.1992) (limits on cross-examination violate defendant’s Sixth Amendment rights only if district court abused its discretion); United States v. McClintock, 748 F.2d 1278, 1289-90 (9th Cir.1984). Reference before the jury to the Ninth Circuit carried an implication that the defendant would be convicted — an implication that many jurors would be sufficiently sophisticated to catch. Such statements should not be made but here the reported comment is not sufficient to taint the verdict. Analogously, the comment on defense counsel’s understanding of honesty did carry a negative implication as to the defendant or his counsel but did not destroy the fairness of the trial.

More generally, a reading of large sections of the transcript gives the impression that the court treated government counsel less brusquely than it treated defense counsel but the difference in treatment is not such as to undermine confidence in the verdict. See Duckett v. Godinez, 67 F.3d 734, 740 (9th Cir.1995) (appellate court will reverse only if there has been an “extremely high level of interference” that “creates a pervasive climate of partiality and unfairness”), ce rt. denied, - U.S. -, 116 S.Ct. 1549 , 134 L.Ed.2d 651 (1996); United States v. Mostella, 802 F.2d 358, 361 (9th Cir.1986) (appellate court should order new trial only if record discloses actual bias or reflects that judge’s comments projected an appearance of advocacy or partiality to jury).

As to one ruling we do have serious misgivings — the ruling limiting to 18 the number of patients the defense might present as witnesses to rebut the government’s theory that Rutgard’s practice was “permeated with fraud.” However, as we subsequently hold in this opinion that the government failed to prove this theory, any arbitrariness in the number of witnesses allowed turns out to be harmless error. A similarly serious but ultimately harmless error occurred in the court’s refusal to permit patient testimony to be offered by the defense at the forfeiture hearing. See United States v. Feldman, 853 F.2d 648, 662 (9th Cir.1988) (district courts must allow defendant to offer evidence regarding the extent of assets subject to forfeiture if the issue is disputed). As we go on to hold that the government did not prove its case for forfeiture, the error was harmless.

We have reviewed Rutgard’s other objections to evidentiary rulings, claim of Brady error, failure to grant immunity to certain witnesses, and exclusion of experts. We find no fatal error.

THE SUFFICIENCY OF THE EVIDENCE

The rules governing appellate review of a challenge to the sufficiency of the evidence are familiar. The jury has been free to make credibility determinations (except in the unusual ease where the witness is incompetent or manifestly incredible) and to draw reasonable inferences from the testimony and exhibits before it. Our sole function is to determine whether any rational juror could have found beyond a reasonable doubt that the defendant was guilty. Jackson v. Virginia, 443 U.S. 307, 319 , 99 S.Ct. 2781, 2789 , 61 L.Ed.2d 560 (1979).

The Cataract Counts. Running through the government’s case on the cataract counts is the way Rutgard used the Brightness Acuity Test or BAT. For 125 years the standard way of testing eyesight has been the eye-chart known as the Snellen, familiar to anyone who has gone to an ophthalmologist. In 1983 there came into commercial production tests to measure the effect of glare on vision. See Thomas C. Prager, Jack T. Holladay, et al., Glare Testing in Cataract Patients: Instrument Evaluation and Identification of Sources of Methodological Error, 15 J. Cataract & Refractive Surgery 149 (1989). The BAT was a device created by Dr. Jack T. Holladay, professor of ophthalmology at the University of Texas in Houston, to measure the effect of glare on vision. It is used in conjunction with the Snellen test in a dark room where the patient reads letters or numbers on lines that are progressively smaller as a way of determining if the patient’s best-corrected eyesight is 20/20 or something worse. “Something worse” for Medicare purposes was 20/40 (meaning the patient can only see at 20 feet what a person of normal eyesight can see at 40 feet) until July, 1990 when Medicare made the minimal standard 20/50.

As a government witness, Dr. Holladay testified that use of the BAT was encouraged by the American Academy of Ophthalmology and common today. He testified that determining if and how glare affects an individual’s living is an essential element in assessing visual functions.

The BAT has three settings — low, medium, and high. These settings • simulate various combinations of illumination and reflectivity. None of the settings correspond to nighttime glare. “Low” simulates bright overhead commercial lighting. “Medium” and “high” simulate Outdoor daylight conditions with overhead sunlight. As both illumination and reflectivity vary, the possible combinations are many. Illustratively, “medium” and “high” both simulate light when a person is standing on a white concrete sidewalk or on a sandy beach looking horizontally at the Snel-len eye chart (i.e., reflectivity is high), but “medium” corresponds to illumination on a cloudy day and “high” corresponds to light on a bright sunny day. In every instance it is the combination of sunlight with the reflectivity of the surrounding surfaces that determines the degree of glare so that there is no single equation between degree of light in San Diego and the amount of glare generated.

As the instruction manual for the BAT prescribes, the tester should move from the BAT off, to low, to medium, to high, giving the patient’s retina time to adjust to the different amounts of light. If a single value has to be chosen, the most appropriate reading would be at the medium setting. The only possible abuse of the BAT known to Dr. Holladay was to use only the high setting and to take the measurement so rapidly that the patient did not adjust to the bright light. Dr. Holladay’s manual on the use of the BAT was found in Rutgard’s home. The jury could infer that Rutgard knew what proper usage was. There was conflicting testimony on whether Rutgard instructed his staff to use the BAT regularly only on high or at least record only the reading irom the high setting. There was also testimony that Rut-gard instructed employees to use whatever BAT setting was necessary for the patient to qualify for medically necessary surgery. If the jury believed that these were his instructions, the jury could infer that a BAT reading without indication of the setting was a high reading based on an improper use of the BAT.

Medicare regulations made no reference to BAT even though the BAT can cause a pa tient’s acuity on the Snellen Test to drop more than six lines. Rutgard argues that, in the absence of a Medicare rule, his use of the BAT was not fraudulent. We agree that he could use the BAT readings. But a reasonable juror could have found a fraudulent pretense of medical necessity if Rutgard deliberately used the BAT improperly. This case is not about violation of the rules of Medicare but about whether Rutgard made representations that were in fact untrue in order to obtain money from Medicare.

Against the background of the testimony about the BAT we examine the cataract counts in terms of individual patients.

Pearl Adams. She had cataracts in both eyes. A government witness, Dr. Russell Edwards, testified that he had not examined her but had examined her chart and found her Snellen test readings to be 20/25 minus 1 in both eyes (minus 1 refers to an inability to read one letter on the line indicated). With BAT the measure was 20/60 minus 2. The chart reflected her complaint of being unable to read newspapers and magazines. In Dr. Edwards’ opinion bifocal glasses would have met her reading problems; her ordinary Snellen scores did not show a need for cataract surgery and she denied having problems with glare. Whether or not surgery was needed depended primarily on the weight given the BAT score. Inferentially, the BAT was set on high to provide false evidence of medical necessity. Combined with Adams’ denial of glare problems, the inference supports the conviction on counts 1, 2, 49, 50, 52 and 53. Rutgard objects that Dr. Edwards cannot pronounce on medical necessity without seeing the patient; but he was competent to give his expert opinion and the jury was free to believe him if it chose to do so. See United States v. Sullivan, 595 F.2d 7, 8 (9th Cir.1979) (jury» not court, is to determine credibility of expert witnesses offering competing views of defendant’s sanity).

John Boyle. In 1987 and 1988 Boyle came to Rutgard with a sense that he needed new glasses; he was having trouble seeing freeway signs at a distance. His left eye had been amblyopic or undeveloped since birth, but his right eye had been 20/20 with glasses. Rutgard found that he had cataracts in both eyes but did not recommend surgery. Boyle returned in 1990 believing he needed bifocals to read small print in the newspapers. He repeated his difficulty in seeing freeway signs at a distance and said that at night, looking at oncoming headlights, “I would see two or three lights instead of one.” His chart showed vision in his right eye as 20/40+ , with “40+ ” being written in over “30”, and vision in his left eye as 20/200. His complaints were listed as “trouble reading fine newspaper print, trouble seeing freeway signs, lights at night like starbursts can’t see driving.”

At trial defense witnesses explained the common practice of translating a patient’s response to a diagnostic procedure into a statement made by the patient. The practice was not fraudulent. Rutgard told Boyle that cataract surgery would improve his right eye and might improve the vision in his left. Boyle testified that surgery had no effect on his left eye. There is no doubt that Boyle had cataracts and that he complained of their effect on his driving. Medicare changed its regulations in July, 1990 to require that vision be at least 20/50 to justify surgery. The “40+ ” entry is not an attempt to meet this standard. Without specific testimony that “30” was altered to 40+ , a jury could not be sure that there had been a fraudulent alteration. A reasonable juror could not conclude beyond a reasonable doubt that the cataract surgery performed on Boyle’s right eye was medically unnecessary but could reach this conclusion as to surgery on his undeveloped left eye. Counts 165 and 166 have not been proved, counts 7, 171 and 172 have been proved.

Robert Church. Church did not testify. His .BAT scores (not identified as to setting) were 20/50 in each eye; the chart reflected alteration by someone at Rutgard’s office from earlier scores of 20/40. Church complained of blurry vision and problems driving, but denied glare problems. In the opinion of government expert Dr. Edwards cataract surgery was not necessary if the BAT scores were based on the high setting of BAT. A reasonable juror could infer that in Rutgard’s practice the high setting was usual and accept Dr. Edwards’ opinion that Church did not have daily living problems that required cataract surgery. Counts 71, 72, 78 and 74 were proved.

Jean Elwood. The government’s evidence was that a form given to Elwood showed a check against the printed form: “Your vision is borderline or below the standard to drive an automobile established by the Department of Motor Vehicles.” Under “History”, her chart reads: “Chief Complaint. Painless progressive loss of vision which is interfering with normal daily activities.” Under “Physician Notes,” there is an indication of her complaining of her vision “interfering with daily activities, including seeing street signs, driving.” Elwood testified that she had not driven since 1980. In presenting the case to the jury the prosecutor said: “This false documentation is why the defendant is guilty of Count 91, for performing that right eye cataract surgery.”

Elwood’s testimony presents a different picture. She came to San Diego in 1985. At the time she possessed a Massachusetts driver’s license. When she went to get a California license she flunked her eye exam. She had not been driving but she could not be licensed to drive because of her eyesight. She came to Rutgard in July, 1991 claiming: “I couldn’t see good. I was—that was the main thing, I couldn’t see good when I was walking or anything. I couldn’t see the television good.” This statement, made during the government’s direct examination, was amplified on cross: she agreed that she “had trouble seeing signs or reading signs.” She responded affirmatively to the question whether she “had trouble in everyday life, just getting around, because of your poor vision.” She admitted having glare problems. She benefitted from the surgery. No reasonable juror could conclude from her testimony that there were material false statements made by Rutgard to justify cataract surgery upon her. She had cataracts. They seriously impaired her living. Their removal led to a significant improvement in her vision. The entry on her chart as to driving was inaccurate but not materially false; it approximated her problems. Counts 13, 91, 92, 94 and 95 have not been proved.

Socorro Manzo. Manzo came to Rutgard already diagnosed as having cataracts in both eyes. Her eyesight was 20/400 in her left eye, even without the BAT. Under “Physician Notes” on her chart it was remarked that she'“complain[ed] of decreasing vision interfering with daily activities, including reading and walking, seeing curbs and steps.... Patient has a history of some macular degeneration, however.” Rutgard told her he could do nothing for her left eye but less than a month later performed cataract surgery on that eye. William Rose, a government witness on billing, testified on the basis of her chart that Manzo’s complaints met the Medicare standard of medical necessity for cataract surgery. Dr. Edwards, also basing his opinion only on Man-zo’s chart, testified that the macular degeneration prevented any reasonable expectation of improvement with cataract surgery. Dr. Tomambe, a defense witness with extensive experience with patients suffering from ma-cular degeneration, testified without reviewing Manzo’s chart that “the macular function is better the more light that gets to it” and that removal of a cataract could improve peripheral vision. With two expert ophthalmologists, neither of whom had seen the patient, disagreeing as to whether the performance of cataract surgery on an eye with a hole at its center is ever medically necessary and with Rose testifying that Manzo’s chart met Medicare standards, no reasonable juror could conclude beyond a reasonable doubt that Rutgard knew the surgery on Socorro Manzo’s left eye was medically unnecessary; the evidence is insufficient to establish fraud. See United States v. Migliaccio, 34 F.3d 1517, 1525 (10th Cir.1994). Counts 10, 20, 115 and 116 are not proved.

Bobbie McKelvey. McKelvey denied making the complaints recorded in her chart about glare problems and problems in driving and seeing street signs. She consented to cataract surgery on her right eye. She was dissatisfied with the result, did not agree to surgery on her left eye, and was suing Rutgard for malpractice at the time of trial. Her insurer was Blue Cross. We are not informed as to what Blue Cross required in order to pay for cataract surgery. Information in the physician’s files cannot be fraud on Blue Cross without such evidence. The jury could believe McKelvey’s denial of the complaints in her file but there is still insufficient proof of false representations by Rut-gard to Blue Cross. Count 26 is not proved.

Marguerite Meador. Meador came to Rutgard already diagnosed by another doctor as having cataracts. Her medium BAT was 20/50 and 20/40; her high BAT was 20/70 in each eye.

At trial she denied the complaints entered on her chart of blurry vision and trouble reading words on television. On cross she testified that she circled complaints in a patient form indicating that she had glare problems at night, problems judging distance, and problems reading the telephone book. Government witness Rose testified that her chart met Medicare’s definition of medical necessity-

The government argued to the jury, “There were no patient complaints in this ease.” The government repeats on this appeal that Meador did not make the complaints “that would have warranted cataract surgery.” The jury could believe her denial of making the complaints entered, but she admitted to making other serious complaints about her vision; her medium BAT scores were fairly high and the high BAT met the Medicare standard; Rose’s expert opinion was that the complaints on her chart justified cataract surgery. In this context the complaints she denied, if eliminated, do not materially affect the diagnosis of cataracts or the medical necessity of surgery; or at least no reasonable juror could conclude beyond a reasonable doubt that medical necessity was fraudulently represented to Medicare. Counts 125, 126, 128 and 129 are not proved.

Ruth Roden. Like Manzo, Roden had ma-cular degeneration in one eye, which also had a cataract. Roden’s best vision in the right eye before surgery was 20/200; after surgery it was 20/300. Reviewing her chart at the trial Dr. Edwards was of the opinion that it was “not very likely that visual acuity would improve significantly with the cataract extraction.” His opinion was based on a view of the benefits of cataract surgery on peripheral vision in conflict with the opinion given by Dr. Tdmambe on the importance of increasing peripheral light. As in Manzo’s case, with no other concrete evidence concerning Roden’s condition a reasonable juror could not be sure beyond a reasonable doubt that Rutgard knew that cataract surgery was unnecessary. Counts 42, 43, 159 and 160, all based on such surgery to Ruth Roden’s right eye, were not proved.

YAG Laser Counts. Rutgard performed YAG laser surgery on Socorro Manzo and Ruth Roden, follow-up operations after their cataract surgeries. Dr. Edwards testified based on the patient charts that the surgeries were not medically necessary because of the macular degeneration. Dr. Tomambe’s refutation went toward the benefits on patients with macular degeneration of cataract surgery, not YAG laser surgery. The government during closing argued that YAG laser surgery affects only central, not peripheral vision, so that it could not help patients with macular degeneration. The jury could reasonably conclude that YAG surgery on these patients was medically unnecessary. Counts 119, 120, 163 and 164 are proved.

Eyelid Procedure Counts. Rutgard performed a large number of eyelid surgeries, often on the same patients he had treated for cataracts. Those surgeries corrected two conditions: ptosis (drooping of the upper eyelid) or ectropion (turning outward of the lower lid). Medicare recorded Rutgard as performing 13.65 percent of the ptosis surgeries done by the approximately 1,000 Specialty 18 Medicare providers in the seven-county area in Southern California serviced by Transamerica in 1990, 24.77 percent in 1991, and 12.42 percent up to April 1992. Rutgard performed 32.9 percent of the ectro-pion surgeries in 1991 and 22.3 in 1992. A number of counts were based on such surgeries.

Depending on the patient involved, the prosecution offered various grounds for finding fraud as to these eyelid surgeries. First, the government argued Rutgard billed for more expensive procedures than he performed. Rutgard performed a form of upper eyelid repair known as a Fasanella-Servat, which took him only a few minutes to do, and billed it as a true blepharoptosis, which takes approximately 45 minutes to perform properly. Rutgard at trial described the Fasanella-Servat as “quick and easy and comfortable for the patients” and on direct examination did not deny that this procedure was what he favored. In the cases where the more complicated operation was billed a reasonable juror could conclude beyond a reasonable doubt that Rutgard had deliberately billed the simple Fasanella-Servat as the more complicated one in order to receive the higher fee it justified. The defense argues that the Medicare billing code for “internal pto-sis” was for the higher fee and that a Fasa-nella-Servat could be billed in good faith to this code number. On this question of billing accuracy, it was for the jury to decide if a billing mistake was honest or fraudulent. With respect to lower lid surgery, Rutgard performed ectropion surgery on both eyes in about ten minutes. Experts testified that true ectropion repair on both eyes takes between 45 minutes and one hour so that Rutgard could not have performed the procedure for which he billed. Here again the jury was entitled to decide whether any billing errors were honest or fraudulent.

Secondly, the government also produced patients who denied the eyelid complaints recorded in their files. The jury was entitled to believe the patients’ testimony and to find the charts falsified.

Thirdly, the government also contended that in particular cases the Polaroid photos showing patients with droopy upper eyelids or droopy lower lids were “staged,” that is, the patients were coached to exaggerate their condition for the photographer so that surgery which was unnecessary would appear to be necessary. The defense disputed this testimony, noting that the close-range Polaroids often produced some distortion and presenting evidence that patients with droopy eyelids sometimes overcompensate to hide their condition and need to be told to relax to get a picture accurately reflecting their condition. The jury was entitled to make its own judgment as to the credibility of the government’s evidence and the defense rebuttal.

In the light of these criteria of evidentiary sufficiency, the following counts were proved:

Counts 3, 55 and 56, ptosis surgery on Pearl Adams, proved by improper billing.

Count 180, note in the file of Edwin George Bebbington as justification for future ptosis surgery. The note read, “My lids are cutting off my vis, when I hold them up I see better.” In his deposition read into the record at trial, Bebbington, an octogenarian who died prior to trial, denied that he thought his eyelids interfered with his vision, but later stated in the deposition, “I remember [Rut-gard] raising [the lid] up and asking about— and I think I could see a little better at that time. I don’t recall particularly if I did, but it seems like I did.” He had not actually had the eyelid operation and could see well now without having had the operation. Bebbing-ton’s deposition, taken as a whole, could have been found by any reasonable juror to have shown him scheduled for an unnecessary surgery.

Counts 8, 174, 175 and 212-215, ptosis surgery on John Boyle, proved by his denial of complaints.

Counts 75 and 76, ectropion surgery on Robert Church, proved by improper billing. Counts 77 and 78, ptosis surgery on him, proved by improper billing.

Counts 9, 79, 80 and 182, ptosis surgery on, and photos of, the right eye of Kevin Collins. The eye was prosthetic; the surgery was cosmetic, not medically necessary. Collins also denied making complaints in his file.

Counts 10, 83-86, 186 and 187, ptosis surgery on, and photos of, Eileen DiPuma, proved by her daughter’s denial of the complaints in the file and by improper billing. Counts 11, 12, 87-90, 188 and 189, ectropion surgery on, and photos of, DiPuma, proved by her daughter’s denial of complaints in the file and by improper billing.

Counts 14, 96, 97 and 190, ptosis surgery on Jean Elwood, proved by her denial of the complaints in her file and by improper billing.

■ Count 47, false diagnosis of ptosis of Charlotte Gerry sent to Aetna, proved by patient’s denial of reported complaints, false visual field test, and staged photograph.

Counts 82, 184 and 185, ptosis surgery on, and photos of, Mary de Herrera, proved by her denial of complaints. (The court described this witness as “very confused” but did not find her incompetent to testify.)

Count 48, false diagnosis of ptosis in both eyes of Brayton Jett-Carroll sent to Aetna, proved by patient’s denial of complaints ascribed to her in the letter.

Counts 21, 117, 118, 198, 199 and 200, ptosis surgery on, and photos and visual field of, Socorro Manzo. She did not testify. Her 75-year-old boyfriend, who had accompanied her to Rutgard’s office, denied she had made the recorded complaints. Rutgard challenges the memory of this witness, but the accuracy of his recall was a question for the jury. The visual field in her file was identified as boilerplate identical with 24 other photocopied fields found in other patients’ charts.

Counts 22, 23, 121, 122 and 201, ptosis surgery on Ralph McClester, proved by his denial of complaints. Counts 24, 25, 123 and 124, ectropion surgery on him, proved by denial of complaints and conflicting evidence about whether the surgery occurred.

Counts 130, 202 and 203, ptosis surgery on, and photo of, Margaret Meador, proved by her denial that the photo was fair and accurate, by her denial of complaints about her eyelids, and by improper billing.

Counts 44, ptosis surgery on Ruth Roden, proved by improper billing.

The following counts were not proved:

Counts 69 and 70, ectropion surgery on Hubert Benjamin, who denied that he complained about his eyelids interfering with his vision but testified to conditions following cataract surgery — tearing, scratching, and redness in both eye — that government witnesses said were common symptoms of an ectropion problem. A reasonable juror could not conclude beyond a reasonable doubt that Benjamin had not effectively complained about the condition of his eyelids.

Vitrectomy Counts. Clark Potter had cataract surgery on both eyes performed by Rutgard; the necessity of this surgery is not disputed. As part of the surgery, Rutgard implanted an intraocular lens. Potter, a physicist with a keen awareness of the physics of vision, found that with the implant he still needed glasses. In the end Rutgard replaced the lens four times before Potter was satisfied. All of the replacements, according to Potter, involved vitrectomies or draining of fluid from the eye. Rutgard billed Medicare only for the vitrectomies, not the lens replacements. When billing clerk Jean Havel asked why, Rutgard replied that he did not have to bill Medicare for everything and that, if he billed for the replacement, Medicare would treat the vitrectomy as part of the replacement and pay a lower fee than if the billing was for a vitrectomy alone without reference to replacement. On the basis of the testimony of billing expert Rose, the jury could find that suppression of the relation of the vitrectomies to the replacements constituted the fraudulent concealment of material facts. Counts 38, 39, 40, 41, 149, 150, 152, 153, 154, 155, 157 and 158, all involving the billing of Potter’s replacement lens, were proved.

The Retrobulbar Injection Counts. Patients Boyle and de Herrera testified that they did not receive injections of medicine for swollen eyes after cataract surgery by means of the formidably long retrobulbar needle. Rutgard at trial testified that he approached with the needle from a patient’s side so that the patient would not see it and be seared. It was for the jury to decide whose memory and whose explanation were credible. Sufficient evidence supported conviction on Counts 81 and 168.

W-5 Modifier Counts. Rutgard billed 81.7 percent of W-5 charges filed by the approximately 1,000 Medicare Specialty 18 providers in the seven-county area in Southern California serviced by Transamerica in 1990, 88.8 percent in 1991, and 96.3 percent in 1992. Government billing expert Rose testified that the W-5 code was justified in Medicare billing during the post-operative follow-up period only if there was “an unusual occurrence,” that is, “if the post-op visit was occasioned by a new visit or something different that was not present for the original event.” In other words, the code was appropriate only if the patient visited the doctor for a problem unrelated to the surgery. Ptosis would be such an event under Medicare rules if that was the reason the patient came to the office. If the doctor discovered the condition during a routine post-op visit, however, it was not billable. As Rose added, this rule “is alien to a lot of physicians.” Adams and Boyle were diagnosed with ptosis during their routine post-op exams and Rutgard billed under W-5. Rutgard argues that the Medicare rules were not clear and that the government never produced the texts of any rule. At trial, however, Rutgard did not seek to have any Medicare rule introduced as “the best evidence”; Rose’s testimony was an adequate basis for the jury to find that a rule regulating the W-5 modifier was in efféct and that Rutgard must have been aware when he used the W-5 code that he was making a false assertion that the post-operative event had been brought to his attention by the patient, as the jury found in its guilty verdict on Counts 54, 169, 170 and 173. Counts 67 and 68 for W-5 billing for Edwin Bebbington were similarly sustained as he received post-op treatment for which his chart indicates he had made no complaint. Count 156 for Clark Potter is sustained because the chart shows treatment during a routine post-op appointment.

The EKG Counts. According to advice that Rose gave Rutgard in 1988 when Rutgard had invited him for recommendations on his billing practices, for Medicare purposes “all services must be billed where they are performed.” Medicare did not pay for EKGs performed at a surgery facility on the day of surgery, treating such measurements as embraced within the facility fee. When bills for such services were rejected, Rutgard began billing the EKGs as if they were performed in his clinic and so obtained the regular reimbursement of approximately $25.00 per EKG.

Rutgard not only makes the argument that no Medicare rules were introduced at trial but also asks us to take judicial notice of the Medicare newsletter published by Trans-ameriea in September, 1987 and the Medicare Billing Manual of March, 1992, both of which have rules as to compensation for EKGs without explicit reference to where the EKGs are performed. It is inappropriate for us to take such notice when these materials could have been introduced at trial by Rut-gard but were not. Moreover, what Rutgard was convicted of was not violation of a particular rule but of making an untrue statement which he found it necessary to make in order to obtain compensation for EKGs from Medicare. The fraud was small but deliberate, material, and cumulatively lucrative. Counts • 51, 93,127,151 and 167 were proved.

The False Entry Counts. These counts overlap with counts already analyzed in terms of the medical condition treated, but present an issue different from medical necessity. They involve patient complaints or tests supporting surgery which were found when the government in execution of a search warrant seized Rutgard’s files. They are Count 183 (visual field for Kevin Collins); Counts 186 and 188 (complaints of Eileen DiPuma); Count 190 (complaints of Jean Elwood); Count 198 (complaints of Socorro Manzo); Count 201 (complaint of Ralph MeClester); Count 202 (complaint of Marguerite Meador); Counts 209 and 210 (complaints of Clark Potter); and Counts 212, 213 and 214 (complaints of John Boyle). The prosecutor presented evidence that these complaints were not made but that they and the visual field were boilerplate added to the patients’ charts, usually to justify surgery for ptosis.

The false entry statute provides:

18 U.S.C. § 1001 . There can be no valid conviction under § 1001 “unless both juris diction and materiality are shown.” United States v. Facchini, 874 F.2d 638, 641 (9th Cir.1989).

Rutgard argues that the patients’ charts in the doctor’s office were not within federal jurisdiction. But Rutgard had an obligation to assure that his Medicare-funded services were “supported by evidence of medical necessity ... as may reasonably be required by a reviewing peer organization,” 42 U.S.C. § 1320c-5(a)(3), and he knew that Medicare’s agent, California Medical Review of Insurance (CMRI), did require documentation of medical necessity by patient complaints. False entries in his files would impair the functions of Medicare by depriving it of the possibility of accurate review of his Medicare billing. The false statements' concerned the authorized functions of Medicare. See Facchini, 874 F.2d at 641 .

It is argued that statements in one’s own files are not statements to a federal agency. The objection is irrelevant. The statute does not only criminalize false statements or representations to a federal agency. It criminalizes each “false entry” made to conceal a material fact. There is no doubt that patient complaints and visual fields are material facts bearing on medical necessity. It is argued that doctors should not be made criminals for inaccurate notetaking. But the statute speaks to fraudulent notations. They are criminal only when, as here, they prevent review of payments made to a physician by the government. The case is governed to a significant degree by United States v. Balk, 706 F.2d 1056, 1059 (9th Cir.1983) (intended use of welders’ certificate was for matter within jurisdiction of Navy). The case is almost on all fours with United States v. Hooper, 596 F.2d 219, 223 (7th Cir.1979) (university’s internal records of student stipends, although not submitted to agency, were within jurisdiction of federal agency funding stipends because of agency’s audit and inspection authority).

Was Rutgard’s Entire Medical Practice a Fraud? In three contexts, the government undertook to prove that Rutgard’s entire practice was a fraud on the insurers: before the jury to prove the § 1957 monetary transaction counts; before the jury to sustain the criminal forfeiture; and before the court at sentencing to establish loss. As the prosecution told the jury, “The scheme to defraud was to treat every eye covered by insurance as a profit center and to do whatever was necessary to milk that eye for as much money as [Rutgard] could possibly get. The scheme was conducted with a total disregard for medical necessity.” In this part of our opinion we review the evidence offered in support of this claim. We then address the three specific contexts in which the evidence is relevant.

■ At the core of the government’s proof of its contention was the testimony of fourteen ex-employees of Rutgard, insiders who testified for the government. Nurse Sigrid Thordin, who worked for Rutgard from October, 1991 to April, 1992, testified that 90 to 95 percent of the photos taken for ptosis and ectropion surgery did not accurately represent the true condition of the patient’s eyelid. Berta To-dor, a technician in Rutgard’s employ from 1987 to 1989, testified that suture removals were routinely misbilled as the removal of a foreign body from the eye, that cortisone injections were misbilled as if rendered to the back of the eyeball, and that the BAT was routinely set on “high” in order to justify surgery. Billing clerk Jean Havel testified about Rutgard’s regular EKG misbillings to Medicare and that the W-5 modifier was “vastly overused”. Nancy Biagioni, who worked as a technician for Rutgard from 1988 until April, 1992, testified that she was told to fill out patient complaint forms by circling certain standard complaints without seeing the patient. She also testified that, as a result of discussing with Rutgard the problem of what to do with patients who needed pre-surgery authorization from an insurer for ptosis surgery but had automatic visual field scores “too good” to get the authorization, she used fake field tests. Jesse Del Valle, Rutgard’s office administrator in 1988, did a review with Rutgard of all patient charts in anticipation of a possible Medicare audit. Del Valle testified: “Any time there was no reason to have done surgery, or any time I saw some surgeries that were not documented, I would put a note on the section that needed to be amended.” Del Valle would then hand the chart to Rutgard, and Rutgard would make the needed entry with his own hands, making up the information that was missing. George Butera, Rutgard’s office administrator from 1988 to 1991, was asked by the prosecutor: “While you were administrator, could you be sure of the accuracy of the information about a patient’s condition and treatment in any chart at Jeff Rutgard’s practice?” Butera answered: “I would say No. I came to believe No.”. Butera testified in particular that visual fields which at Rut-gard’s direction he inserted in patients’ charts which CMRI was to review were in fact “fictitious,” i.e. not based on actual observation but drawn to justify ptosis surgery that had already occurred. He also testified that Rutgard paid $10,000 to a former employee who threatened to report Rutgard to Medicare.

These questionable practices took place in an ophthalmologic practice where around 10,-000 patient visits occurred each year, where patients were often herded through examinations, where frequently Rutgard himself spent no more than two minutes examining a patient, and where profitmaking was the focus of the physician.

When he took the witness stand in his own defense, Jeffrey Rutgard had a lot to explain. He made blanket denials of fraud as to particular practices, but he failed to explain a great deal. The government offered to waive the three-hour time limit on direct examination. The defense declined the offer and did not even take Rutgard on redirect. Rutgard’s reticence is not comparable to invocation of the privilege of self-incrimination, an invocation that the prosecution cannot comment on. Griffin v. California, 380 U.S. 609, 615 , 85 S.Ct. 1229, 1233 , 14 L.Ed.2d 106 (1965). He freely testified and left the jury to draw inferences from what he did not say. The government on this appeal contends that he waived any challenge to the sufficiency of the evidence by testifying and cites United States v. Ledesma, 499 F.2d 36, 43 (9th Cir.1974), for this conclusion. The government’s argument goes too far. What Ledesma holds is that the jury is entitled to take into account deficiencies observed in the defendant’s own testimony as evidence against him. Here, for example, the government seized on Rutgard’s testimony that he advised against Anthony Colarusso having eyelid surgery. The government produced Colarusso’s chart showing him complaining that his droopy eyelids interfered with his reading, driving and other daily activities and that he was scheduled by Rutgard for surgery to correct ptosis of both eyes. The government had a basis for inviting the jury to believe that Rutgard had perjured himself in his own defense and so could not be trusted in his general denials.

The government did not, however, provide sufficient evidence that Rutgard’s entire medical practice was an insurance fraud. Three convergent reasons lead to that conclusion. First, the government did not charge or prove that his medical practice was a criminal enterprise of the kind condemned by RICO, 18 U.S.C. §§ 1961-68 ; nor did the government charge or prove a conspiracy to defraud the government or that the fourteen ex-employees who testified or any other employees were co-conspirators with Rutgard. It may be that the exigencies of prosecuting Rutgard necessitated a soft approach with the employees. Whatever the reason, the ex-employee witnesses presented themselves at trial as doing honest work, only occasionally doing recording or billing against their honest judgment because of Rutgard’s overriding directions. The inescapable inference is that a fair amount of recording and billing in the Rutgard practice was performed without fraud.

Take Butera, for example, whose damning remark about coming to believe that none of the files were accurate was relied on by the government before the jury and is relied on by the government in this appeal. Butera testified that in 1990 he, as administrator, was proud of the practice and ready to let CMRI review the files. His comment on inaccuracies is careful not to call them fraudulent. Even the time he came to doubt the accuracy of the files is unfixed. Fraud must be proved by more specific testimony as to time and extent. Butera’s testimony about the fake visual fields supporting ptosis surgery is of this character; his general “I came to believe” is not. It is not without signifi- canee that at trial both the defense and the government treated a number of entries in patient charts as reliable. Del Valle testified to fictitious entries when they were needed, not to fictitious entries in every file. Not every entry justifying treatment was a fraud.

Second, there are portions of Rutgard’s practice where no fraud was suggested. He had some small number of cash patients, who were not insured. He did some work with glaucoma — how much is not clear, but reference is made at trial to diagnosis of glaucoma and to treatment of glaucoma by laser surgery. It would be surprising in any ophthalmology practice where a high percentage of patients are over 65 that glaucoma was not a complaint of some patients. In 1991,19 percent of the visits of Medicare patients nationally seeking eye care were related to glaucoma. Ellwein et ah, supra at 1732. But no fraud was charged as to glaucoma patients. Some of his patients also must have had retina problems, which nationally accounted for 14% of Medicare eye visits in 1991. Id. But fraud as to the retina was not charged. It was the burden undertaken by the government, not the defense, to show that 100 percent of Rutgard’s medical practice was fraudulent.

Third, no fraud was shown on the counts that we reverse with this opinion. In addition, the government did not charge or show fraud as to the cataract surgeries performed on its witnesses de Herrera, MeClester and Potter; nor did the government show fraud as to the cataract surgeries performed on Rutgard patients Bickerton, Buros, Colarus-so, Daniels, Izzarelli, Laing, Mikkelson, Pop-kin, Schweitzer, See, and Shindell, all of whom testified for the defense with their charts in evidence.

As the government observes, patient satisfaction was not an issue in this case. Rut-gard was on trial for insurance fraud, not malpractice. Nonetheless, it is improbable that all of Rutgard’s satisfied cataract patients were persons who did not need cataract surgery. The actually-proved instances of fraudulent pretense of medical necessity for cataract surgery are a tiny fraction of a practice that did thousands of cataract surgeries. Given the national figures for cataract surgeries paid for by Medicare in 1991, Ellwein et al., supra at 1734, it appears that almost one in two of Medicare beneficiaries of eye care (almost one in four of all Medicare beneficiaries) need such surgery, unless the unfair and unsupported belief is entertained that Medicare eye practice everywhere is fraudulent. With a high probability that many persons over 65 do in fact need the removal of cataracts, it is unlikely that all of Rutgard’s cataract practice was a fraud. The court did not permit the defense to offer more than 18 patient witnesses. Those who did testify show that an undetermined percentage of Rutgard’s patients were satisfied and their cataract surgeries were not performed as fraud upon an insurer.

For these three reasons, converging to show that Rutgard’s practice was not simply or wholly a scheme to defraud insurers, we hold that the government failed to prove beyond a reasonable doubt or even by a preponderance of the evidence that Rut-gard’s entire practice was a fraud and that all its proceeds were the fruit of fraud.

On appeal the government rephrases its jury argument that the entire practice was a fraud to say that the practice was “permeated with fraud.” The government urges us to uphold the jury verdict on that basis. The phrase “permeated with fraud” is a term used to sustain a search warrant in circumstances where business records are so mingled that the fraudulent and the honest cannot be distinguished prior to the search. See, e.g., Center Art Galleries-Hawaii, Inc. v. United States, 875 F.2d 747, 750-51 (9th Cir.1989). A finding of “permeated with fraud” is sufficient to establish probable cause in the non-adversary context of an application for a search warrant. See United States v. Hendershot, 614 F.2d 648, 654 (9th Cir.1980). The conclusory phrase is too loose and indefinite a term to sustain a criminal conviction beyond a reasonable doubt or to constitute a finding by the preponderance of the evidence absent specific supporting evidence. To reach a finding of preponderance there has to be a comparison of the evidence, pro and con; the determination to be made by the trier of fact is different in kind from the probability assessment made by a magis trate issuing a warrant. United States v. Ventresca, 380 U.S. 102, 107-08 , 85 S.Ct. 741, 745-46 , 13 L.Ed.2d 684 (1965). To prove its ease here the government had the burden of showing that there were no medically necessary, properly billed surgeries. The government did not meet this burden.

In the light of this analysis we put the evidence in the three contexts in which the government relied on its theory that every eye was a profit center for Rutgard “with a total disregard for medical necessity.”

The Monetary Transactions Count. In May, 1992 after the government’s search of Rutgard’s office, his wife Linda, at his direction, made two wire transfers to the National Westminster Bank on the Isle of Man, one of $5,629,220.74 on May 5 and one of $1,935,220.48 on May 6. The transfers were made from the Rutgard Family Trust, whose accounts were at the Imperial Trust Company in San Diego.

The government’s accounting expert noted $15.8 million paid by Medicare to Rutgard entities between the beginning of 1988. and April, 1992. He testified that Rutgard deposited $3,754,056 derived from these entities into the family trust account during this period and deposited $1.9 million in municipal bonds so that a total of $5,654,056 entered this account from October 29, 1990, when Rutgard opened the account with a personal cheek for $560,000, to May, 1992. He testified that the balance of the account came from municipal bonds delivered to the trust account at unspecified times. On appeal the government argues that the jury could take into account the first year, 1987, when insurance fraud began, and come to the conclusion that the entire amount transferred by wire transfers on May 5 and 6 was the proceeds of insurance fraud.

To prove this contention the government advanced the theory that all of Rutgard’s practice was a fraud. If the government had succeeded in this proof, it would have properly convicted him of the monetary transaction counts. But we have just determined that the government’s proof was deficient. The government, as we have also held, established Rutgard’s guilt of particular counts of fraud. The proceeds of that fraud come to over $46,000. Can the convictions under § 1957 be sustained on the theory that at least $20,000 of fruits of fraud were incorporated in the wire transfers of May 6 and 7? See United States v. Adams, 74 F.3d 1093, 1100-01 (11th Cir.1996) (holding that at least $10,000 of each transfer charged must be criminally derived property). To answer that question requires a consideration of the terms and purpose of § 1957 and a close look at the companion statute, § 1956.

These statutes govern monetary transactions in criminally derived property. The standard money laundering statute is 18 U.S.C. § 1956 . The other statute, the one under which Rutgard was convicted, is 18 U.S.C. § 1957 . In construing § 1956, we referred to § 1957 as “a companion money laundering statute,” United States v. Garcia, 37 F.3d 1359, 1365 (9th Cir.1994), cert. denied, - U.S. -, 115 S.Ct. 1699 , 131 L.Ed.2d 562 (1995), without having occasion to mark the differences between the two. See also United States v. Lopez, 104 F.3d 1149 (9th Cir.1997) (per curiam). We now have occasion to mark those differences.

Section 1956, alone at issue in Garcia , bears the title “Laundering of monetary instruments.” It punishes by imprisonment of up to 20 years and a fine a defendant who:

18 U.S.C. § 1956 (a)(1). For present purposes, five elements of § 1956 differentiate it from § 1957, the statute at issue here — its title, its requirement of intent, its broad reference to “the property involved,” its satisfaction by a transaction that “in part” accomplishes the design, and its requirement that the intent be to commit another crime or to hide the fruits of a crime already committed.

. Section 1957 has a different heading: “Engaging in monetary transactions in property derived from specified unlawful activity.” It punishes by up to ten years’ imprisonment and a fine anyone who:

18 U.S.C. § 1957 (a). The description of the crime does not speak to the attempt to cleanse dirty money by putting it in a clean form and so disguising it. This statute applies to the most open, above-board transaction. See 18 U.S.C. § 1957 (f)(1) (broadly defining “monetary transaction”). The intent to commit á crime or the design of concealing criminal fruits is eliminated. These differences make violation of § 1957 easier to prove. But also eliminated are references to “the property involved” and the satisfaction of the statute by a design that “in part” accomplishes the intended result. These differences indicate that proof of violation of § 1957 may be more difficult.

Section 1957 could apply to any transaction by a criminal with his bank. Two years after its enactment an amendment was necessary to provide that the term “monetary transaction” does not include “any transaction necessary to preserve a person’s right to representation as guaranteed by the sixth amendment to the Constitution.” Pub.L. 100-690, § 6182 , 102 Stat. 4181 , 4354 (1988). Without the amendment a drug dealer’s check to his lawyer might have constituted a new federal . felony.

Section 1957 was enacted as a tool in the war against drugs. See Anti-Drug Abuse Act of 1986, Pub.L. 99-570, Subtitle H, 100 Stat. 3207 (1986). It is a powerful tool because it makes any dealing with a bank potentially a trap for the drug dealer or any other defendant who has a hoard of criminal cash derived from the specified crimes. If he makes a “deposit, withdrawal, transfer or exchange” with this cash, he commits the crime; he’s forced to commit another felony if he wants to use a bank. This draconian law, so powerful by its elimination of criminal intent, freezes the proceeds of specific ciimes out of the banking system. As long as the underlying crime has been completed and the defendant “possesses” the funds at the time of deposit, the proceeds cannot enter the banking system without a new crime being committed. United States v. Savage, 67 F.3d 1435, 1442-48 (9th Cir. 1995), cert. denied, - U.S. -, 116 S.Ct. 964 , 133 L.Ed.2d 885 (1996). A type of regulatory crime has been created where criminal intent is not an element. See Morissette v. United States, 342 U.S. 246, 252-56 , 72 S.Ct. 240, 244-46 , 96 L.Ed. 288 (1952). Such a powerful instrument of criminal justice should not be expanded by judicial invention or ingenuity. We “should not enlarge the reach of enacted crimes by constituting them from anything less than the incriminating components contemplated by the words used in the statute.”- Id. at 263 , 72 S.Ct. at 250 .

For these reasons we do not find helpful in interpreting § 1957 the cases applying § 1956, which speaks of design “in whole or in part” and of “a financial transaction involving property.” See, e.g., Garcia, 37 F.3d at 1364-66 ; United States v. Jackson, 983 F.2d 757, 765 (7th Cir.1993). Other circuits, however, have used the § 1956 precedents to eliminate any tracing of funds in a § 1957 case. See United States v. Moore, 27 F.3d 969, 976-77 (4th Cir.1994); United States v. Johnson, 971 F.2d 562 (10th Cir. 1992). They have reasoned that otherwise § 1957 could be defeated by a criminal mingling innocently-obtained funds with his ill-gotten moneys. E.g., Moore, 27 F.3d at 977 .

This reasoning rests on the fungibility of money in a bank. That fungibility, destroying the specific identity of any particular funds, makes the commingling of innocent funds with criminal funds an obvious way to hide the criminal funds. If § 1956 required tracing of specific funds, it could be wholly frustrated by commingling. For that reason, the statute not only proscribes any transaction whose purpose is to hide criminal funds but reaches any funds “involved” in the transaction. Neither the same reasoning nor the same language is present in § 1957, the statute here applied.

The monetary transaction statute cannot be made wholly ineffective by commingling. To prevail, the government need show only a single $10,000 deposit of criminally-derived proceeds. Any innocent money already in the account, or later deposited, cannot wipe out the crime committed by the deposit- of criminally-derived proceeds. Commingling with innocent funds can defeat application of the statute to a withdrawal of less than the total funds in the account, but ordinarily that fact presents no problem to the government which, if it has proof of a deposit of $10,000 of criminally-derived funds, can succeed by charging the deposit as the crime; or the government may prevail by showing that all the funds in the account are the proceeds of crime. E.g., Savage, 67 F.3d at 1440-43 . Commingling will frustrate the statute if criminal deposits have been kept under $10,-000. But that is the way the statute is written, to catch only large transfers. Moreover, if the criminal intent was to hide criminal proceeds, as would presumably be the case any time criminally derived cash was deposited with innocently derived funds to hide its identity, § 1956 can kick in and the depositor of amounts under $10,000 will be guilty of a § 1956 crime.

The government did not take its possible course of charging Rutgard with deposits of over $10,000 of fraudulent proceeds. The government had the means of doing so because its accounting expert identified the large deposits Rutgard made. But as Rut-gard was neither charged nor convicted of deposits in violation of § 1957, we cannot uphold his convictions on that basis.

Rutgard’s convictions may be upheld if he transferred out of the account all the funds that were in it or if there was a rule or presumption that, once criminally-derived funds were deposited, any transfer from the account would be presumed to involve them for the purpose of applying § 1957. Rutgard did not transfer all the funds in the family trust account, however. The government showed that the account held $8.5 million on April 2,1992 and $13,901 on July 2,1992, the dates of the quarterly bank statements. But so far as evidence at trial goes, more than $46,000 remained in the account after the May 5 and 6 transfers. These transfers therefore did not necessarily transfer the $46,000 of fraudulent proceeds.

The alternative way of sustaining the convictions depends on a presumption, which the Fourth Circuit created in Moore, 27 F.3d at 976-77 , but which we decline to create. The statute does not create a presumption that any transfer of cash in an account tainted by the presence of a small amount of fraudulent proceeds must be a transfer of these proceeds. Unlike § 1956, § 1957 does not cover any funds “involved.” To create such a presumption in order to sustain a conviction under § 1957 would be to multiply many times the power of that draconian law. It would be an essay in judicial lawmaking, not an application of the statute. As the government did not prove that any fraudulently-derived proceeds left the account on May 5 or May 6, 1992, the monetary transfer counts, Counts 216 and 217, were not proved beyond a reasonable doubt.

The Criminal Forfeiture. The evidence for the verdict of criminal forfeiture under 18 U.S.C. § 982 (a)(1) was the evidence submitted to the jury on the counts on which it had already convicted. The government argued that all of the money involved in the § 1957 monetary transfers was forfeit. The government added, “There is no evidence for any other source of income in the Rutgard Family Trust other than the proceeds of Jeff Rutgard’s eye business.” The defense objected that this argument shifted the burden of proof to the defense. The objection was overruled. The district court erred in this ruling. The government, not the defendant, had the burden of showing the sources of all the moneys in the account, although the burden here was lowered to a preponderance of the evidence. See United States v. Hernandez-Escarsega, 886 F.2d 1560, 1576-77 (9th Cir.1989); United States v. Myers, 21 F.3d 826, 829 (8th Cir.1994), cert. denied, - U.S. -, 115 S.Ct. 742 , 130 L.Ed.2d 643 (1995).

The district court, as we have already noted, also erred in prohibiting patient defense witnesses at this stage of the proceedings. Neither ruling, however, is ultimately harmful because the government stood or fell on the same evidence on which it tried the monetary transfer counts. The government’s position was that all of Rutgard’s practice was a fraud. The government failed to prove this proposition in terms of the monetary transaction counts and, for the three reasons already stated, failed to prove this proposition to establish that the $7.5 million dollars transferred by wire were subject to criminal forfeiture.

THE SENTENCE

The Sentencing Enhancements. That Rutgard had a leadership role in the crimes of which he was convicted was established with the proof of their commission. U.S.S.G. § 3Bl.l(a). That he took advantage of vulnerable victims and abused trust was established because, in a professional medical practice, trust between patient and physician is essential and because the government as insurer depends upon the honesty of the doctor and is easily taken advantage of if the doctor is not honest. U.S.S.G. §§ 3A1.1, 3B1.3; United States v. Stewart, 33 F.3d 764, 770 (7th Cir.1994) (upholding victim vulnerability enhancement where elderly targets of defendant’s scheme did not suffer financial loss but funeral homes did); United States v. Bachynsky, 949 F.2d 722, 735-86 (5th Cir.1991) (finding that patients were vulnerable victims of defendant’s scheme to defraud Department of Defense and insurance companies by submitting false diagnoses and opining that Department of Defense and insurance companies probably were); United States v. Adam, 70 F.3d 776, 782 (4th Cir.1995) (holding that physician making claims for Medicare enjoys a position of trust allowing him to commit difficult-to-detect wrongs constituting abuse of trust). The enhancement for obstruction of justice was based on four examples of “knowingly false testimony” pointed to by the government at sentencing and adopted by the district court. See United States v. Robinson, 63 F.3d 889, 892 (9th Cir.1995) (sentencing enhancement for obstruction of justice through perjury requires that defendant gave “(1) false testimony, (2) on a material matter, (3) with willful intent”). The judgment that Rutgard committed perjury in testifying was a factual determination by the district court which we have no reason to disturb. See United States v. Shrestha, 86 F.3d 935, 938 (9th Cir.1996) (factual findings during sentencing are reviewed for clear error).

The enhancement for loss falls into a different category. Under the Guidelines, loss generally need only be established by a preponderance of evidence, not beyond reasonable doubt. United States v. Watts, - U.S. -, 117 S.Ct. 633, 637 , 136 L.Ed.2d 554 (1997) (per curiam); United States v. Restrepo, 946 F.2d 654, 661 (9th Cir.1991) (en banc). The loss caused by a scheme of mail fraud is properly reflected in sentencing. United States v. Fine, 975 F.2d 596, 600 (9th Cir.1992) (en banc). . At sentencing, basing itself upon the evidence it heard during the 79 days of time, the court found Rutgard’s practice “permeated with fraud.” On that basis it found a loss of more than $10 million and added 15 points to his score for sentencing. U.S.S.G. § 2Fl.l(b)(l)(P).

As we held earlier in this opinion, “permeated with fraud” is a conclusion that is too indefinite to sustain a criminal conviction absent specific supporting evidence. It is also too indefinite and eonclusory to support a sentence. In modern federal practice sentencing is nearly as critical a stage for the defendant as the trial itself. The standard of proof is relaxed but proof there must be. Restrepo, 946 F.2d at 661 . A global “permeated with fraud” conclusion is not a judgment based on specific evidence.

We have held that where misrepresented grapes were delivered the fraudulent broker must still be given credit for the value of the grapes he did deliver. United States v. Licciardi, 30 F.3d 1127, 1134 (9th Cir. 1994). Analogously, Rutgard must be given credit for the medical services that he rendered that were justified by medical necessity. As always, the burden is on the government to establish what services were not medically necessary. The global estimate of loss made by the district court cannot stand.

The Sentence of Restitution. We review the legality of the restitution order de novo. United States v. Rice, 38 F.3d 1536, 1540 (9th Cir.1994). The district court ordered restitution (less the sums forfeited) of all of Rutgard’s billings to the government and private insurers for the period covered by the indictment. As to fraud prior to November 29,1990, the sentence violates the rule laid down by the Supreme Court that under the Victim and Witness Protection Act (VWPA), 18 U.S.C. § 3663 , restitution may only be ordered for “the loss caused by the specific conduct that is the basis of the offense of conviction.” Hughey v. United States, 495 U.S. 411, 413 , 110 S.Ct. 1979, 1981 , 109 L.Ed.2d 408 (1990); United States v. Baker, 25 F.3d 1452, 1456 (9th Cir.1994).

Effective November 29,1990, the VWPA was amended to provide that “a victim” under the act was “any person directly harmed by the defendant’s criminal conduct in the course of the scheme, conspiracy, or pattern.” 18 U.S.C. § 3663 (a)(2). The amendment had the purpose and effect of undermining Hughey . After the amendment, restitution may be ordered for losses to persons harmed in the course of the defendant’s scheme even beyond the counts of conviction. See United States v. DeSalvo, 41 F.3d 505, 515 (9th Cir.1994). At the same time this substantive change in the criminal law cannot be applied retroactively without violation of the ex post facto clause of the United States Constitution. Id. Consequently, the sentence of restitution cannot stand. On resentencing, restitution may be ordered as to losses arising from specific counts of conviction prior to November 29, 1990 and thereafter as to that portion of the defendant’s conduct proved at trial to have directly harmed an insurer.

CONCLUSION

We AFFIRM Rutgard’s conviction of Counts 1-3, 7-12, 14, 21-25, 38-41, 44, 47-56, 67-68, 71-90, 93, 96-97, 117-124, 127, 130, 149-158,163-164,167-175,180,182-190,198-203, 209-210 and 212-215. We REVERSE his convictions of Counts 13, 19, 20, 26, 42, 43, 69, 70, 91, 92, 94, 95, 115, 116, 125, 126, 128, 129, 159, 160, 165, 166, 216 and 217 and the judgment of criminal forfeiture. We VACATE his sentence and REMAND for resen-tencing.