ORDER

The mandate issued on November 7, 2002, is hereby recalled for the purposes of clarifying the amended opinion.

The Government’s unopposed motion for clarification of amended opinion and for expedited consideration is GRANTED. The Amended Opinion, 2002 WL 193023 (9th Cir.(Cal.)), filed in this matter on October 30, 2002, is further amended and the further amended opinion is filed concurrently herewith.

The mandate shall issue forthwith.

OPINION

Roy Blueford appeals his jury conviction of being a felon in possession of a firearm. Although Blueford raises several challenges on appeal, we need address only one, namely the allegation that the prosecutor engaged in misconduct that tainted the trial. We conclude that the prosecu tion’s actions in asking the jury to infer that Blueford had fabricated his alibi in certain telephone calls with witnesses in the weeks just before the trial, when in fact the government had evidence contradicting some of its assertions, requires that we reverse and remand to the district court for a new trial.

I. Facts and Proceedings Below

In response to reports of drug trafficking, Oakland, California, Police Officer Eric Richholt conducted an undercover surveillance operation in February 1999. While on duty, he observed two men engaged in conversation approximately 75 feet from his surveillance vehicle. Already familiar with the defendant in this case, Richholt identified one of the men as Roy Blueford and the other as Raheem Gibson, a friend of Blueford’s. After a few minutes, Richholt observed the man he believed to be Blueford take a black pistol from underneath his coat, show it to the other man, and then place it behind a nearby fence.

When the two men left the area on foot, Richholt directed other officers to arrest Blueford. They found him minutes later amongst a group of people outside a nearby apartment. Gibson was not among them. The officers also found a gun matching Richholt’s description behind the fence, as Richholt had indicated.

In March 1999, the grand jury returned a single-count indictment charging Blue-ford with being a felon in possession of a firearm. Before trial, Blueford informed the government that he would be presenting an alibi defense, claiming that he had not spoken with Gibson nor been at the location where the gun was placed on the day in question but instead had spent the entire day in the company of three friends: Jumoke Clay and the Fountain brothers, Mike and Orlando. Pursuant to Federal Rule of Criminal Procedure 12.1(a), Blue-ford provided the government with notice that he intended to rely on an alibi defense, and, later, with a list of potential alibi witnesses, among them Clay and Orlando Fountain.

Believing that “Blueford was colluding with his friends to create this alibi,” the government directed the Alameda North County Jail, where Blueford was incarcerated, to record his telephone calls to thirty different phone numbers. Recording commenced on December 18, 1999; the government told the district court that it collected cassettes of the recorded calls on December 29, and that there were no telephone calls thereafter.

Blueford’s trial began on January 3, 2000. The government rested on the following day, January 4, and Blueford called his first witness that afternoon. Later that evening, the government informed Blueford’s counsel for the first time that Blueford’s telephone calls had been recorded and provided her with fifteen cassettes of those conversations at 8:25 p.m., and another nineteen cassettes at 10:10 p.m. Each cassette is capable of storing thirty minutes of recorded material, but none was used to its full capacity. Ten of them were unintelligible, although they were not marked as such. The district court found that the government informed defense counsel that some of the second batch of tapes were “unclear.” Defense counsel says in her sworn declaration that the government’s lawyer “told me that there were additional tapes that were unintelligible or inaudible and that he was providing me with only those tapes that were intelligible”; there is no finding to the contrary. The tapes contained a total of approximately two-and-one-half hours of audible conversation. Defense counsel, working with two others in her office, listened to some of the first batch of tapes that evening.

The following morning, Blueford objected to the timing of the tapes’ disclosure pursuant to Federal Rule of Criminal Procedure 16(a)(1) and moved to prohibit the government from using the tapes at trial. The district court ruled that any statements by Blueford on the tapes would be excluded, but that statements by persons with whom Blueford was conversing were not discoverable under Rule 16 and could therefore be used by the government to impeach Blueford’s witnesses. The prosecutor stated his intent to use the tapes for that purpose: “I am probably going to use them as impeachment material by using the statements of alibi witnesses that they’ve made to Mr. Blueford.”

Defense counsel then requested a continuance so that she could review the cassettes. The district judge responded: “You can listen to them after court today.” The judge also stated that the material from the tapes would not be admissible until the following day, January 6. As it happened, the defense team did not finish listening to the cassettes until several days later, after Blueford’s trial and conviction. According to defense counsel, it took almost 30 hours to review the tapes. Although there were, it turned out, only two- and-one-half hours of intelligible conversation on them, the recorded and intelligible parts were, as noted, not separately marked, so it was not possible simply to listen to the intelligible parts.

On January 5, Blueford testified that he had been with Jumoke Clay and the Fountain brothers at the time he allegedly committed the charged offense. During cross-examination, the government cast doubt on Blueford’s alibi by suggesting that it had been fabricated in telephone conversations in the weeks immediately before the trial. Toward that end, the government questioned Blueford extensively about the frequency of his telephone calls to potential alibi witnesses, emphasizing that Blueford had many more phone conversations with those individuals in the period just before the trial than he had in the preceding months. In doing so, the prosecutor asked both about many specific conversations and about whether there were, in general, more conversations with the potential alibi witnesses in the last week or two — that is, from December 22 on — than previously. For example, the questions regarding Blueford’s calls to Jumoke Clay’s number included:

During a break in the cross-examination, Blueford renewed his motion for a continuance to review the tapes, arguing that the government was basing questions on “the fruit of the discovery violations.” The court denied the motion, stating that the cross-examination to that point had relied only on the “trap-and-trace” records — ie., the register of the date, length, and destination of outgoing calls made by Blue-ford- — not the content of any particular conversation as revealed by the tapes. The court suggested that defense counsel “have someone on [her] staff listening to the tapes while we’re in session.”

The next defense witness was Orlando Fountain, who testified that Blueford had been with him and others- — but not Gibson — on the day of the alleged offense. The government cross-examined Fountain about the frequency and length of his phone conversations with Blueford, again asking about specific calls and emphasizing the high volume of calls in the period just before the trial. For example, cross-examining Orlando Fountain, the prosecutor asked:

In the end, the prosecutor did not introduce any of the tapes into evidence. The trap-and-trace records were not introduced either (so as not to reveal to the jury that the defendant was in custody), but a stipulation as to what those records revealed concerning calls to Orlando Fountain’s home was. That stipulation showed no such calls until November 20 (Blueford had testified that he did not have Fountain’s new telephone number until about November), and a number of such calls thereafter, including one on December 23, five on December 28, and six between December 29 and January 1. It appears that the stipulation included twenty-six of the twenty-eight calls made between November 20 and January 1 that lasted more than 20 seconds; we are not told why two were left out, if they were, or when they occurred.

In its closing argument, the government asked the jury to infer from the recent flurry of telephone conversations that Blueford and Fountain had fabricated an alibi:

In total, the trial took three days. The jury, after deliberating over the course of three days (part of January 6, all of January 7, and part of January 10) and asking for readbacks of Officer Richholt’s and Blueford’s testimony, convicted Blueford of being a felon in possession of a firearm. Blueford then moved for a new trial, claiming that the late disclosure of the tapes had impaired his defense. Blueford’s lawyer also declared that, in listening to the tapes after the trial, she learned that they contained statements by Blueford inconsistent with the inference of fabricated alibis, including a query as to why Raheem Gibson “would lie on me” by testifying for the government, and telling a potential alibi witness “[A]ll you got to do is tell the truth.”

The government opposed the motion for a new trial on the ground that the cassettes were not discoverable material, and that there was enough time to review them. In support of its first contention, the government explained that the tapes contained no discussion of the events on the day of the offense, and no discussion concerning the fabrication of an alibi:

The district court judge stated more than once that she was quite surprised to hear this:

The district court denied the motion for new trial and sentenced Blueford to 63 months’ imprisonment. This appeal followed.

After the notice of appeal was filed, the government—at the court’s invitation—filed a narrative Proposed Order Re: Defendant’s Motion for a New Trial. Among other things, the Proposed Order states that “Blueford has not pointed to any relevant statements in the tapes,” and “because Blueford was aware that his calls were being monitored, the taping of his calls turned out to be futile.” In a footnote, the Proposed Order asserts that “[t]he fact that there is[sic] no tapes after December 29, 2000 does not indicate, as Blueford suggests, that the monitoring seized [sic] on December 29, 2000, rather the lack of tapes suggests that Blueford made no phone calls after December 29, 2000.” The district court issued an opinion explaining its denial of the motion for a new trial on June 6, 2000.

II. Prosecutorial Misconduct

Blueford argues that the prosecutor engaged in prejudicial misconduct by asking the jury to infer from the trap-and-trace records that Blueford’s conversations at the end of December involved “concocting an alibi,” when the government knew that (1) some of the particular conversations that were the basis for the inference did not contain any such fabrication; (2) none of the recorded conversations contained any discussion that supported the general inference the jury was asked to draw; and (3) there was material on the tapes directly contradicting the inference, including Blueford’s statement that a potential witness should just tell the truth and another questioning why a potential adverse witness would lie. We agree that the prosecutor’s actions at trial — exacerbated, we would add, by the various implausible explanations the government has given after trial and on appeal of its actions concerning the tapes and the trap- and-trace records — violated defendant’s right to a fair trial, were prejudicial, and require a new trial.

A. General Principles: “It’s the easiest thing in the world for people trained in the adversarial ethic to think a prosecutor’s job is simply to win.” United States v. Kojayan, 8 F.3d 1315, 1324 (9th Cir.1993). It is not. An attorney for the government is a “representative not of an ordinary party to a controversy, but of a sovereignty whose obligation to govern impartially is as compelling as its obligation to govern at all; and whose interest, therefore, in a criminal prosecution is not that it shall win a case, but that justice shall be done.” Berger v. United States, 295 U.S. 78, 88 , 55 S.Ct. 629 , 79 L.Ed. 1314 (1935), overruled on other grounds, Stirone v. United States, 361 U.S. 212 , 80 S.Ct. 270 , 4 L.Ed.2d 252 (1960). Put differently: “The prosecutor’s job isn’t just to win, but to win fairly, staying well within the rules.” Kojayan, 8 F.3d at 1323 .

It is certainly within the bounds of fair advocacy for a prosecutor, like any lawyer, to ask the jury to draw inferences from the evidence that the prosecutor believes in good faith might be true. But it is decidedly improper for the government to propound inferences that it knows to be false, or has very strong reason to doubt, particularly when it refuses to acknowledge the error afterwards to either the trial court or this court and instead offers far-fetched explanations of its actions. Id. at 1318-19 ; see also id. at 1321 (the difference between a lawyer “ask[ing] the jury to infer only things that he believed in good faith might be true” and making “factual assertions he well knew were untrue” is “the difference between fair advocacy and misconduct”); United States v. Udechukwu, 11 F.3d 1101, 1106 (1st Cir.1993) (“[i]t is improper to imply reliance on a fact that the prosecutor knows to be untrue”); United States v. Valentine, 820 F.2d 565, 566 (2nd Cir.1987) (finding prejudicial misconduct where “the prosecutor misrepresented, at least implicitly, the substance of the testimony of several grand jury witnesses”).

B. The Tapes: The government had on tape many conversations conducted during the last two weeks before trial. The prosecutor had already reviewed the tapes when he asked Blueford and Fountain about specific recent conversations and later argued to the jury that the phone calls “confirm” that Blueford’s alibi was false. It is now undisputed that the tapes contained virtually no discussion of Blueford’s trial or the underlying offense, and. — in the government’s own words- — -“there’s no statement [on the tapes] that says something like, ‘This is what I want you to testify to at trial,’ or something like that.”

Despite his knowledge that there was nothing on the tapes to support the assertion that Blueford had asked Fountain to concoct an alibi, the prosecutor asserted in closing argument that the phone calls “confirm[ed]” his theory that Fountain “concocted that story now for the trial.” Absent some explanation, then, it appears that the prosecutor was asking the jury to infer one or more facts that he either knew to be false or, at least, could not have believed might be true, given that he had specific evidence indicating the contrary.

C. The Government’s Explanations: The government’s explanations of this apparent inconsistency have varied:

1. The Earlier Calls Theory: The government’s answering brief had only one response to the charge that it had asked the jury to draw an inference that it (the prosecutor) knew to be false. That brief pointed out that, as demonstrated by the stipulation based on the trap-and-trace records, Blueford and Orlando Fountain had spoken by telephone many times in November and early December, before taping had commenced, and argued that there is good reason to believe an alibi was concocted at that time.

But that is decidedly not the theory that the prosecutor implied by his questions to Blueford and Fountain, nor the theory argued to the jury. Over and over again, the prosecutor emphasized the frequency of phone conversations that took place in late December, when recording was underway: “[Y]our amount of talking to [Fountain] increased quite a bit in the last week and a half, two weeks; “especially in the last two weeks, you’ve called [Fountain] maybe seven or eight times;” “on December 28th, for example, you actually called him one, two, three, four times;” “so you spoke to him quite a bit recently in the last week or two before your trial;” “especially in the last two weeks, you’ve spoken to Jumoke quite a bit, haven’t you;” “between December 25 and December 81, ... that’s more than you’ve spoken to him in the past couple months, is that right;” you [ie., Fountain] spoke to [Blueford] on December 29, 1999 is that right? ... You also spoke to him on December 28th four times ... is that right?” In its summation, the government emphasized that Blueford and Fountain had recently conversed a great deal, “especially in the week and a half before this trial.” That there is another, somewhat similar, factual inference the government could have asked the jury to draw that was not contradicted by the tapes does not make the prosecution’s invitation to the jury to infer a fact that is contradicted accurate or proper. See Valentine, 820 F.2d at 569-71 (prosecutor improperly implied a factual pattern that might have been proper and relevant had the government’s theory of the case been otherwise).

2. The Partial Recording or Distinct Calls Theories: At oral argument, the government took a new tack in explaining why the inference pressed on the jury through cross-examination and closing argument was not inconsistent with what the government knew about the phone conversations: The government now noted that the trap- and-trace records contain some calls in the weeks just before trial that were not recorded at all or were recorded unintelligibly. Most extravagantly, the government maintained that the phone calls made in late December that were the subject of its cross-examination were an entirely different set of calls from the ones captured on the cassettes, as the length of calls in the summary of the recorded calls and the stipulation concerning the trap-and-trace records do not match. In the government’s words, the cross-examinations were only about the trap-and-trace, “not about the recorded phone calls.... They don’t overlap.” As the fabrication could have occurred in unrecorded calls, the government argued to us, the prosecutor had no reason to doubt the suggestion to the jury regarding fabrication of an alibi.

Taking, first, the less extreme version of this explanation, based on the correct observation that there were unrecorded telephone calls during the relevant time period, we conclude that this thesis does not provide a proper basis for the prosecutor’s cross-examination and jury argument, for two related reasons.

First, the government not only asked the jury to draw a general inference from the overall pattern of calls, but also, in cross-examining Blueford and Fountain, pointed to particular conversations as supporting the inference. Some of those conversations must have been the same ones that were recorded, yet the tapes are flatly inconsistent with the fact the prosecutor asked the jury to infer.

The government’s submission concerning the recorded conversations shows four conversations between Fountain and Blue-ford on December 28; the stipulation concerning the trap-and trace records shows five calls on that day; and the prosecutor asked both Blueford' — twice—and Fountain about four conversations that day, spending more time on that day than any other. None of the recorded conversations between Blueford and Fountain on December 28 support the fabrication inference the jury was asked to draw, in part on the basis of those conversations. Even if there was one call on that day that was unrecorded or unintelligibly recorded, the prosecutor could not properly imply that the pattern of calls that day — their repetition and timing — “confirm” that an alibi was fabricated, when he knew that the four of the five conversations contained no such discussions.

Similarly, the cross-examination of Blue-ford concerning the Clay conversations focused in large part on the calls on Christmas Day, December 25. There were three recorded calls that day, and the prosecutor asked about three, based on the trap-and-trace records, just before asking, “So you’d agree with me, in the last week, you’ve spoken to Jumoke quite a bit.” The prosecutor also asked about four calls to Clay on December 29; there was one — very long (over twenty-one minutes) — call to Clay that day on the tape, which did not include any support for an alibi fabrication inference. Clay did not testify' — according to defense counsel, because she feared that there was material on the tape that would contradict Clay’s alibi testimony — and the inference the jury was asked to draw in the prosecutor’s oral argument was limited to fabrication of an alibi with the witness who did testify, Orlando Fountain. Nonetheless, Blueford told the jury that Clay was with him the whole time, so the insinuation that he was discussing his alibi with Clay in the December telephone calls still had resonance, as support for the inference concerning the alleged Fountain fabrication calls during the same time period.

Second, while it is, of course, possible that the particular conversations in late December that were for some reason not recorded were the ones on which alibi fabrication occurred, the fact remains that the recorded conversations over the emphasized time period contained no attempts to fabricate an alibi, and that the government knew that. Additionally, after the trial the government represented to the district court its belief that the monitoring had failed to uncover fabrication of the alibi because the defendant was aware that the recording was going on. If that was the government’s understanding, then the basis for any reasonable belief that the inexplicably unrecorded or unintelligible calls were different in character from those that were intelligibly recorded is all the weaker.

In short, the general inference the jury was asked to draw was against all the available confirming evidence, and against the government’s later explanation about why the recordings showed no alibi fabrication. That being the case, the cross-examinations and argument to the jury concerning the overall pattern of calls during the immediate pre-trial time period were at least seriously misleading. See Valentine, 820 F.2d at 571 (misconduct where “the prosecutor’s presentation gave the trial jury an unfair and inaccurate impression that documentary evidence and oral testimony uniformly substantiated the government’s ... theory,” yet there was some grand jury testimony that the government knew about that was inconsistent with its “consistent pattern” argument.)

As to the government’s more extreme suggestion that none of the recorded calls were those identified on the trap-and-trace records, and vice versa:

We find this explanation unhelpful to the government’s attempt to justify its behavior in this case, for several reasons. To begin with, we are reluctant to consider positions newly minted for oral argument. Rodriguez v. Marshall, 125 F.3d 739, 743 (9th Cir.1997). While we ordinarily welcome clarifications of the record, the government’s novel presentation of the case is anything but clarifying, for it introduces an unresolved mystery: How could it come about that there were two distinct sets of phone conversations in late December, one (recorded) in which Blueford says nothing relevant to his impending trial, and the other (unrecorded) in which Blueford and his friends are busy concocting an alibi? The government has provided no explanation of why the two sets of telephone calls should be entirely distinct, or why the recorded set would have no alibi fabrication while the trap-and-trace set would.

Further, the government’s assertion at oral argument is contradicted by the government’s representations to the district court concerning the relationship between the trap-and-trace records and the recordings. The government told the district court, concerning the process by which prisoners’ phone calls are recorded by the Alameda North County Jail, that:

The prosecutor also explained that when recording is requested, the Jail “ha[s] a large mechanism that tapes everything,” Although the mechanism eventually records over the tapes, the government asked the Jail to retain all the tapes of Blueford’s calls to the relevant phone numbers after a certain date.

As we read these statements, they are inconsistent with the government’s belated suggestion that the tape-recorded and pen-register identified calls were entirely different from one another. At the very least, while some of the recordings may have gone astray, all the calls that were recorded must also appear on the trap- and-trace records. As all but two of Fountain’s trap-and trace identified calls of any length were included in the stipulation, there could not be more than two recorded Fountain calls left out of the stipulated calls on which the inferred fabrication was based. Similarly, it is clear from the record that the prosecutor cross-examined Blueford concerning his calls with Clay based on the trap-and-trace records, and that he purported to be questioning about all the calls on certain days, including Christmas. Thus, at least some of the calls on December 28 to Fountain and December 25 to Clay were both among the recorded calls that did not support the fabrication allegation and among the trap- and-trace identified calls about which the prosecutor questioned the witnesses in support of the inference he later propounded to the jury.

It is true that the recorded and trap- and-trace calls on the same day are noted as being of different lengths. For example, although the questions to Blueford regarding calls to Clay on December 25, based on the trap-and-trace records, mentioned calls of thirteen, ten, and nine minutes, the government’s summary of the taped calls attributes to that day calls of nine and one-half minutes, fifteen seconds, and sixteen and one-half minutes; there are similar discrepancies in the taped and pen-register summaries for the December 28 calls to Orlando Fountain. We are not sure why that is. The discrepancies could reflect (1) differences in how the call lengths were recorded ; (2) the possibility that where the recordings are shorter (that is not always the case), some of the conversation may have been lost; or (3) the fact that, as both the defendant and the government explained, Blueford was using calls to Clay and Fountain to connect him with third parties, such as his mother and girlfriend; any call is supposed to be terminated at that point but that does not always happen, and it is possible that the trap-and-trace records and tape recordings are affected differently once another line is connected. Whether the discrepancy can be traced to one of those three reasons or to some other explanation, the government’s current observation that the duration information does not match — with no explanation as to why that might be so— cannot overcome its clear representation to the district court that all calls were' identified on the trap-and-trace records, so that the recorded calls must be among the calls so identified. Further,- it is also worth noting that the government apparently did not notice the duration discrepancy until oral argument, so that discrepancy cannot be the basis for a good faith belief during the trial in the truth of the suggested inference.

We note that we are particularly leery of the government’s most recent explanation of its behavior at trial concerning the late-December phone calls because there is a pattern in this case of after-the-fact explanations for prosecutorial conduct that collapse upon examination. One such instance is the attempted reconciliation in the government’s brief, discussed above, which gives way once the actual representations at trial are examined. Another appears in the government’s proposed order filed after the motion for a new trial was denied: In attempting to demonstrate that there was no bad faith on the part of the government in providing the tapes to Blueford in the middle of trial, that proposed order stated that the government “disclosed the tapes as soon as they were finished monitoring the calls,” and, in a footnote, accounted for the period between December 29 and January 4, the day the tapes were disclosed, by telling the court that “the fact that there are no tapes after December 29, 2000 does not indicate ... that the monitoring [ceased] on December 29, 2000, rather the lack of taping simply suggests that Blueford made no phone calls after December 29, 2000.” But we know that Blueford did make phone calls after December 29, 2000. The prosecutor knew it as well, because he asked Blueford in cross-examination questions concerning phone calls made on December 31 and January 1 and entered into a stipulation based on the trap-and-trace records including calls on December 31 and January 1. The government also indicated to the trial court that the tapes were collected on December 29 and did not mention any later collection of tapes, thus further calling into doubt its explanation of why it did not turn over the tapes to the defense until several days later. Positing at oral argument two distinct sets of telephone calls strikes us as a similar stab-in-the-dark explanation that cannot be squared with what the district court was told before making its rulings.

We conclude that the prosecutor at trial improperly asked the jury to infer that the pattern of calls in late December demonstrated that Blueford was using the calls to concoct an alibi with prospective witnesses.

III. Harmless Error

As a general matter, although there are exceptions, see Kojayan, 8 F.3d at 1325 , establishing that there has been prosecutorial misconduct is not in and of itself sufficient to merit reversal of a conviction. Rather, if there was an adequate objection, ór if objection was not required, the question is whether the misconduct was harmless. United States v. Sanchez, 176 F.3d 1214, 1218 (9th Cir.1999); Fed. Rule Crim. P. 51.

If, on the other hand, there should have been a contemporaneous objection but there was not, we review for plain error, and will vacate a conviction only when “there is (1) error, (2) that was clear or obvious, (3) that affected substantial rights, and (4) that seriously affected the fairness, integrity, or public reputation of the judicial proceedings.” United States v. Vences, 169 F.3d 611, 613 (9th Cir.1999) (internal quotation marks omitted); see also Fed.R.Crim.P. 52.

Here, both determining whether review should be for harmless error or plain error and assessing the prejudice caused by the prosecutor’s conduct require that we take into account the timing and circumstances of the release of the tapes to defense counsel. Doing so, we conclude that by releasing the tapes mid-trial even though they were available several days earlier and by implying that there was material on them inconsistent with Blueford’s alibi defense, the prosecution both exacerbated the prejudice caused by its conduct before the jury and unfairly deprived the defense of any real opportunity to object to the cross-examination and jury argument. Where a party has “no opportunity to object to a ruling or order,” he may not be prejudiced for failing to do so. Fed. Rule Crim P. 51; Valentine, 820 F.2d at 571 .

The tapes through December 29 were produced at the end of the government’s case, although, according to the government’s own account, they were available several days before. While the government represents that the tapes could not be released earlier because doing so would have interfered with the monitoring, as far as the record shows, no calls were recorded after December 29. As noted above, while the government tried to rescue its representation by suggesting that the monitoring did continue after the 29th but produced no calls because there weren’t any, that explanation cannot be squared with the government’s own questions and stipulation at trial.

The government suggests that the defense could have and should have listened to the tapes once they were handed over. Had the defense done so, the defendant could have made the objection that the prosecutor was implying that there was fabrication discussed on some of the calls when there was not.

We view the question whether the defendant lacked fair opportunity to object as quite distinct from whether or not the district court should have granted a continuance, a question we do not address. Without the continuance, however, defense counsel was put to the choice of conducting the usual trial preparations or listening to the tapes on the afternoon and evening of Wednesday, January 5. Although the district court found that there was sufficient time to listen to the tapes as there was only two-and-one-half hours of intelligible material, that conclusion does not account for the difficulty of extracting the intelligible material from the much longer tapes. One cannot know for sure that there is nothing on a tape without listening to it, so the fact that many of the tapes turned out to be partially blank or unintelligible does not mean that the defense had no reason to listen to them. Nor, as the government argues, does the fact that Blueford himself was a party to the conversations mean that he recalls them in detail or can recount the duration of the various conversations, nor did he know which conversations were captured on tape. More importantly, defense counsel herself (or through staff) had to listen to the tapes if she was to respond to them at trial, no matter what her client told them about his conversations.

Further, with regard to the issue directly before us, whatever else may have been on the tapes, the defense had good reason to proceed on the premise that the tapes would include some support for the prosecution’s insinuation that there was discussion on the tapes concerning concocting an alibi. Indeed, both defense counsel and the district judge assumed, because the government’s behavior concerning the tapes led them to assume, that the tapes contained information consistent with the inference that the government, through its questioning and argument, asked the jury to draw.

After the trial, when it was advantageous to do so in order to prove that the tapes were not discoverable in the first place and that no prejudice had occurred due to their late release, the government told the court, and defense counsel, that there was no such information on the tapes. When arguing during the trial, however, concerning the admissibility of the tapes, the prosecutor — who had already listened to the tapes — conveyed quite a different impression, stating that he was probably going to use the recordings to impeach any alibi witnesses. He made only technical arguments against suppression — -that the disclosure was timely, that they were not relevant to the government’s case-in-chief, that the statements would be admissible as impeachment even if there was a discovery violation. Additionally, the prosecution, by making the very insinuations in question, also made the implicit representation that the tapes would back them up. Defense counsel was not obliged to assume the government was suggesting that the jury make inferences that were contradicted by the pertinent tapes, although that turned out to be the case.

We note as well that all of the questioning regarding the telephone calls took place on January 5. The district court, in denying the second continuance request, ruled that no evidence based on the tapes could be introduced until the next day, January 6. As the district court recognized in so ruling, introduction of the tapes before then would not have been fair, as defense counsel had not yet had any opportunity to listen to many of the tapes. So it is quite clear that the defense could not possibly have objected contemporaneously to the questioning concerning the telephone calls, as that questioning occurred before she could have listened to all the tapes. Any objection would therefore have been after much of the damage had been done.

Then, defense counsel, faced on January 5 with the decision whether to interrupt her trial preparation to listen carefully to the tapes or instead, assume they were harmful to the alibi defense and proceed accordingly, did the latter. The combined result of the prosecution’s later-than-necessary disclosure and lack of candor at the pertinent time about what was on the tapes was that the defense counsel, having made a reasonable decision about how to deal with scarce resources of time and staff, did not in fact have the information she would have needed to object during trial to the prosecution’s insinuations, or to put on evidence to counter those insinuations. Under these circumstances, it is fair to conclude that the defendant did not have an opportunity to object that the inference the prosecution was asking the jury to draw was improper, and that we should review for harmless error.

Having done so, we conclude that “[h]ad the prosecutor done his job ... the verdict could well have been different.” Kojayan, 8 F.3d at 1323 . Blueford’s only defense was that he was not where the police officers said he was, and that he had friends who could confirm that. While the prosecution presented other bases for disbelieving Blueford and Fountain, there was the need to account for the fact that if they were lying, they were lying somewhat (although not entirely) consistently, so they must have “concocted” the stories together; the question was how and when. The government offered the jury a very specific answer to that question — in phone calls between them, mostly in the week and a half before trial.

Had the jury known that the prosecutor’s explanation regarding an obvious and critical issue was not supported by the only available tapes of the pertinent telephone calls, it might well have rejected the proffered inferences and concluded that the government had not proven its case beyond a reasonable doubt. “Evidence matters, closing argument matters; statements from the prosecutor matter a great deal.” Id. at 1323 . As it was, the jury did deliberate over a considerable period of time. Further, the jury asked for read-backs of Richholt’s and Blueford’s testimony while it was deliberating, so it evidently did not regard the case as an easy one. That being so, we cannot say that one or more jurors would not have found a reasonable doubt as to Blueford’s guilt had the trial been free from prejudicial conduct.

CONCLUSION

We conclude that the government in this case failed, both at trial and thereafter, to fulfill its responsibility to “discharge its responsibilities fairly, consistent with due process,” id., and that its failure to do so was not harmless. We therefore REVERSE and REMAND for a new trial.

. In its Petition for Rehearing, the government indicates that this information was incorrect, and that in fact calls were recorded thereafter. That the government erred in its earlier representations is not a reason to remand for further factual development now. . This information, as will appear, is pertinent to the length of time it took to listen to the tapes. Had defense counsel been informed that she was being given unintelligible tapes, then it might have been reasonable for the government to expect that she would not need much time to listen to them. The same is not true, however, of tapes that are only "unclear.” . The government states in its brief that defense counsel listened to all of the first fifteen tapes in two hours the night she received them, and that she knew that all of the conversations with Orlando Fountain were on those first tapes. Defense counsel's statements during the trial are ambiguous in this regard. Her more specific post-trial declaration says only that she and two other people were able to listen to “some” of the first set of tapes that first evening, nor is there any way she would have known, without listening to them, whether there were more calls to Orlando Fountain on the second set of cassettes. .Federal Rule of Criminal Procedure 16(a)(1) provides in relevant part: Upon request of a defendant the government must disclose to the defendant and make available for inspection, copying, or photographing: any relevant written or recorded statements made by the defendant, or copies thereof, within the possession, custody, or control of the government, the existence of which is known, or by the exercise of due diligence may become known, to the attorney for the government. On July 12, 1999, Blueford requested from the government disclosure of materials and statements covered under Rule 16(a)(1) and Brady v. Maryland, 373 U.S. 83 , 83 S.Ct. 1194 , 10 L.Ed.2d 215 (1973). . This latter statement was made to support the representation that “the government disclosed the tapes as soon as they were finished monitoring the calls,” even though (1) the stipulation included calls identified through "trap-and-trace” after December 29; and (2) the record shows that the prosecution collected tapes from the jail on December 29 but did not turn them over to the defendant’s counsel until January 4. . We do not rely in this opinion on those statements. . We deny Blueford’s motion to strike the government's post-argument submissions, which consisted of selected portions of the excerpts of record purporting to reveal the discrepancies. Because these materials were previously included in the excerpts of record, we may not strike them. . It is unclear why the questioning, based on the trap-and-trace records, concerned only four conversations, while the stipulation also based on those records mentions five. The reason may be that one of the stipulated calls was quite short, as was one of the recorded calls, so the prosecutor assumed that there was no actual conversation. If that is correct, then it is still true that there was at most one substantive unrecorded (or unintelligible) call that day. . We note that as to the tape recordings, the only duration information in the record is attached to a declaration of Eric Bauer, a federal officer who "listened to all clear conversations on these tapes” and put together a chart that reflects, among other things, the "approximate length of the tape.” . As the government indicated to the district judge: "What many prisoners do, and what it appears Mr. Blueford is doing, is ... make a three-way phone call. In other words, ... they make a collect phone call to a friend, and then they ask the friend to call a third party.” (As Blueford explained, the reason he did this was that some of his relatives and friends had "blocked” phones, and the jail telephones do not connect to such phones, presumably because monitoring such phones is not possible.) Obviously, if the duration discrepancies are traceable to this habit, the discrepancies cannot support the government's theory that Blueford and Fountain or Blueford and Clay were concocting an alibi in the unrecorded minutes of conversations, because Blueford was not actually on the phone with Fountain or Clay during those periods. . Sanchez did not decide, nor do we, whether harmless error due to prosecutorial conduct of the kind here is properly analyzed under the standard for constitutional trial error established in Chapman v. California, 386 U.S. 18 , 87 S.Ct. 824 , 17 L.Ed.2d 705 (1967), or some intermediate standard, see United States v. Bagley, 473 U.S. 667 , 105 S.Ct. 3375 , 87 L.Ed.2d 481 (1985). Here, the prosecuto-rial conduct was not harmless under the standard of harmless error review most favorable to the government, so there is no need to address the question whether some other standard applies. . We do not address, here or elsewhere, the question whether the district court ruled correctly on the Brady and Rule 16 discovery questions or on the continuance requests. Rather, we assume that she did, as that assumption does not affect our conclusion. . When so ruling, the district court did not take into account the possibility that the government was already implying by its questioning a fabrication that had not occurred on the tapes, as it believed from the government's statements that the tapes established otherwise.