OPINION

We must decide whether we have subject matter jurisdiction over a corporate qui tarn action under the False Claims Act that accuses a competing company of committing fraud against the United States.

I

This saga began in the early 1990s, when Aventis Pharma S.A., (“Aventis”) a French pharmaceutical enterprise, developed a plan to sell the drug enoxaparin in the United States under the brand name Lovenox. Enoxaparin is a heparin-based blood thinner that was patented in Europe in the 1980s as European Patent 40,144 (“EP144”). Eventually, European authorities invalidated EP144 as lacking novelty.

A

In June 1991, Aventis applied to the U.S. Patent and Trademark Office (“USP-TO”) for a patent to cover a supposedly new version of enoxaparin (“618 Product”). Aventis filed a New Drug Application (“NDA”) with the Food and Drug Administration (“FDA”) a month later.

The initial versions of the NDA were rejected in part as being “obvious in view of the prior art,” including EP144. Aventis acknowledged that EP144 was prior art, but claimed that the 618 Product was a different drug from the drug produced by following EP144 (“EP144 Product”). After several subsequent rejections, Aventis submitted a revised NDA in 1993 supported by the expert declaration of Dr. Andre Uzan, a French scientist. Dr. Uzan claimed that data in Example 6 of the patent application, included in the revised NDA, “clearly demonstrate^]” that the 618 Product has a longer half-life relative to the EP144 Product, and thus was novel.

The USPTO then approved the revised application and issued Patent Number 5,389,618 (“618 Patent”) for the 618 Product in 1995, in large part because of Dr. Uzan’s statements regarding Example 6. However, it turned out that Dr. Uzan’s sworn declarations were incorrect. Dr. Uzan had based his claim of a difference by comparing a 40mg dose of the 618 Product with a 60mg dose of the EP144 Product, but had not disclosed that different dosages were used in the USPTO submissions. A dose-for-dose comparison would have shown that the difference between the products was “not statistically significant.” This error went undetected for years. Meanwhile, after the 618 Patent was issued, Aventis listed it in the FDA’s Orange Book, which invokes the protections of the Hatch-Waxman Act. 21 U.S.C. § 355(j)(5)(B)(iii) (requiring a generic drug competitor to certify that the patent is either not infringed or invalid).

1

Amphastar Pharmaceuticals, Inc., (“Am-phastar”) a U.S.-based generic pharmaceutical firm founded in 1999 by Yong Feng Zhang, decided that enoxaparin would be a good prospect for a generic product, believing it might be possible to develop a generic version by utilizing information from EP144. Development work began in 2000 and included experiments performed from 2000 to 2003 (“Pre-Litigation Experiments”). The exact nature of the Pre-Liti-gation Experiments is disputed, but they led to Amphastar’s being able to manufacture a generic enoxaparin. Amphastar then submitted an Abbreviated New Drug Application (“ANDA”) to the FDA on March 4, 2003.

Aventis responded to Amphastar’s ANDA by bringing suit five months later, alleging that Amphastar and Teva Pharmaceuticals were infringing on the 618 Patent. Amphastar filed its Amended Answer and Counterclaim and made no mention of the Example 6 error or the Pre-Litigation Experiments.

After discovery was underway, Amphas-tar filed a motion to amend its Answer and Counterclaim, stating, “based on facts recently developed in this litigation, Amphas-tar has added new affirmative defenses and several antitrust claims to its answer and counterclaims.” Amphastar’s additions included an affirmative defense alleging inequitable conduct by Aventis in obtaining the 618 Patent. After Aventis filed a motion to strike the amendments, Amphastar responded by asserting: “As to the unen-forceability of the ‘618 patent based on inequitable conduct, all of the facts and evidence upon which this affirmative defense is based are in the possession and control of Aventis.” Amphastar did not mention the Pre-Litigation Experiments when it listed the evidence on which it based its inequitable conduct defense. Am-phastar also alleged antitrust violations.

2

Amphastar eventually prevailed in its patent infringement litigation battle with Aventis. By the end of the litigation the Federal Circuit had affirmed the district court’s conclusions that: (1) Aventis made material nondisclosures and false representations to the USPTO regarding the 618 Patent application and (2) Aventis intended to deceive the USPTO. Amphastar was thus able to produce generic enoxapa-rin without fear of infringement—the 618 Patent was “found to be unenforceable on the ground of inequitable conduct.” However, Amphastar’s antitrust counterclaim was dismissed.

B

Amphastar then brought this qui tam action against Aventis. Amphastar duly sent the statutorily required letter to the United States Department of Justice, notifying it of its intent to bring this action. Amphastar then filed its sealed complaint on January 7, 2009, alleging that by committing fraud against the USPTO, Aventis obtained an illegal monopoly over enoxapa-rin and then knowingly overcharged the United States. By committing such fraud, Amphastar further alleged, Aventis knew the patent was invalid and thus knowingly defrauded the United States. The complaint was unsealed and the case proceeded in October 2011, after the government declined to join the litigation.

1

On November 14, 2012, the district court ruled that Amphastar’s complaint was based on public disclosures, and therefore Amphastar had to show it was the “original source ... of the information on which the allegations are based.” § 3730(e)(4). Thereafter, Aventis filed a motion for summary judgment contending Amphastar had failed to meet the original source exception to the jurisdictional bar. To be an original source of information underlying the complaint, Amphastar needed to meet the pre-suit disclosure and “direct and independent knowledge” requirements.

Aventis first argued that Amphastar had failed to disclose enough information in its pre-suit disclosure letter. The district court denied Aventis’s summary judgment motion on this issue in a May 2014 ruling, but certified its ruling for interlocutory appeal, which Aventis brought in No. 14-56382.

2

The district court then proceeded to the “direct and independent knowledge” issue. It held a four-day evidentiary hearing to determine if Amphastar had direct and independent knowledge of any of the information underlying its allegations. The district court made various factual findings: Amphastar had no knowledge of Example 6 being false before the Aventis I litigation began; Zhang (Amphastar’s CEO) was not a credible witness; Amphas-tar had engaged in inconsistent actions regarding whether it had known of the Example 6 error before the Aventis I litigation, and such inconsistent actions undermined its argument. The district court also concluded that Amphastar did not produce enoxaparin by following EP144 in its Pre-Litigation Experiments, but copied key information from Patent 618. For all these reasons the district court held that Amphastar had neither direct nor independent knowledge of any of the elements of fraud supposedly committed by Aventis. The district court then dismissed the case for lack of jurisdiction. Aventis moved for attorneys’ fees, which motion was denied.

Amphastar timely appealed the dismissal of its qui tarn suit. Aventis timely cross-appealed the denial of attorneys’ fees.

II

On appeal, Amphastar contends that the district court erred in dismissing its qui tam suit for lack of jurisdiction. It argues specifically that the district court erred in concluding that the allegations of fraud had b.een publicly disclosed and further erred in concluding that Amphastar was not “an original source” of Aventis’s allegedly fraudulent actions. 31 U.S.C. § 3730 (e)(4).

The relevant portions of Section 3730(e)(4) of the False Claims Act state:

If Amphastar’s allegations are based on publicly disclosed information, the jurisdictional bar applies unless Am-phastar is an “original source.” To be an original source, Amphastar must demonstrate that: (1) it had direct and independent knowledge of the information, on which it based its allegations, and (2) it satisfied the pre-suit disclosure requirements by providing the necessary information to the government before filing an action. Amphastar “bears the burden of establishing subject matter jurisdiction by a preponderance of the evidence.” Alcan, 197 F.3d at 1018.

A

Was the public disclosure bar triggered?

1

Transactions or allegations are disclosed if they can be found in pleadings or other public filings. Hagood v. Sonoma Cnty. Water Agency, 81 F.3d 1465 , 1474 n.13 (9th Cir. 1996). Our circuit interprets “allegations” to refer to direct claims of fraud, and, we interpret “transaction to refer to facts from which fraud can be inferred.” United States ex rel. Mateski v. Raytheon Co., 816 F.3d 565, 571 (9th Cir. 20.16). We have adopted the reasoning of United States ex rel. Springfield Terminal Railway Co. v. Quinn, which states that when a critical mass of the underlying facts or of the allegations in the qui tam complaint have been disclosed prior to the qui tam complaint being filed, the public disclosure bar applies. Mateski, 816 F.3d at 571 (citing 14 F.3d 645, 653-54 (D.C. Cir. 1994)).,

The Mateski court explained that if “X + Y = Z, Z represents the allegation of fraud and X and Y represent its essential elements.” Id. (quoting United States ex rel. Found. Aiding the Elderly v. Horizon W., Inc., 265 F.3d 1011, 1015 (9th Cir. 2001), amended on denial of reh’g, 275 F.3d 1189 (9th Cir. 2001)). If the fraud allegations [Z] are disclosed, the bar ap plies. Id. If enough of the underlying facts making up the elements of fraud are disclosed, the bar applies. Id. “[I]n a fraud case, X and Y inevitably stand for but two elements:” the misrepresented facts [X] and the true facts [Y]. Id. (quoting Found. Aiding, 265 F.3d at 1015 ).

2

Amphastar argues that the allegations of fraud were never publicly disclosed before it filed this qui tam action. It concedes that general allegations of fraud were made in its Aventis I Amended Answer and Counterclaim, but says these past allegations never mentioned any false claims submitted to or paid by the federal government and state governments. Instead, its current allegations concern the fact that false claims were submitted to one or another government, allegations which were never made in any prior publicly disclosed document.

However, “[a]n allegation need not include an express reference to the [False Claims Act]” for the public disclosure bar to apply. Alcan, 197 F.3d at 1019. Nor does a disclosed allegation need to contain every specific detail to constitute a disclosure. Id. (rejecting an argument the relator’s new allegations are meaningfully different because the old allegations “did not mention the [False Claims Act] and did not include details of the violations”). Al-can involved a situation where the publicly disclosed allegations “did not mention any overcharging, false-invoicing ... or any other specific fraud on the government.” Id. Like Alcan, this is not a case where the prior allegations concerned “a very different problem” than the qui tam allegations. See Mateski, 816 F.3d at 574 (discussing the “substantially similar” spectrum) (quoting Found. Aiding, 265 F.3d at 1016 ).

Amphastar’s Amended Answer and Counterclaim in the 2003 Patent Infringement case made nearly identical allegations to those made here. It accused Aven-tis of obtaining an invalid patent due to misrepresentations. It also claimed that, despite knowing about these misrepresentations, “Aventis has attempted to maintain or obtain a monopoly over others” by doing things such as improperly listing the Patent in the Orange Book. Finally, it alleged that “Aventis has wrongfully derived, and will continue to wrongfully derive, income and profits from this conduct.” The only new allegation Amphastar makes in the instant case is that the government also bought the drug while Aven-tis held its illegal monopoly, but this is an obvious inference based on the publicly disclosed allegations. The allegations in this case are so “substantially similar” to the prior allegations that we are satisfied the public disclosure bar applies. See Mateski, 816 F.3d at 573-74 .

3

While the fraud allegations being disclosed are enough to trigger the bar, Amphastar also failed to demonstrate that both the misrepresented facts and the true facts were not disclosed. This is also enough to trigger the bar on its own. The misrepresented facts presented to the government (via the USPTO) are that the 618 Product differed from the EP144 Product because the half-life of the products was different, and thus Aventis held a valid patent—and, the true facts (actual state of the world) are that Example 6 was false and thus Aventis held an invalid patent, and this invalid patent allowed Aventis unlawfully to charge monopoly prices to customers. In this case “the information publicly disclosed ... contained the material facts underlying [the relator’s] allegation of-fraud.” A-1 Ambulance Serv., Inc. v. California, 202 F.3d 1238 , 1245 (9th Cir. 2000); see also Found. Aiding, 265 F.3d at 1015-17 (asking if the true state of the world was disclosed (i.e. the defendant’s substandard care) and if the misrepresented facts were disclosed (i.e. the false statements to the government)).

The evidence that the misrepresented facts and the true facts were actually disclosed is overwhelming. The qui tarn complaint alleges that: (1) Aventis intentionally made material misrepresentations to the USPTO regarding whether the 618 Product was novel relative to prior art, (2) such misrepresentations allowed it fraudulently to obtain the 618 Patent, (3) Aventis then improperly listed the 618 Patent in the FDA Orange Book, and (4) such listing allowed Aventis to overcharge the government. The 2004 allegations that were publicly disclosed included all of these facts, except, instead of mentioning overcharging the government, they mentioned overcharging in general. While the Amended Answer and Counterclaim did not specifically mention Example 6, it did allege material misrepresentations regarding “studies of the pharmacological properties of prior art compounds, including half life.”

4

Amphastar also argues that the district court erred because it recited: “The ‘public disclosure’ standard is not intended to be difficult to meet. Hagood, 81 F.3d at 1476 n.18 (noting that courts often treat the ‘based upon public disclosure’ step as a ‘quick trigger’ to arrive at the more exacting ‘original source’ inquiry).” Amphastar Pharm., Inc. v. Aventis Pharma S.A., No. EDCV-09-0023, 2012 WL 5512466 , at *7 (C.D. Cal. Nov. 14, 2012) (‘Aventis V’). Mateski specifically rejected a defendant’s argument that the “quick trigger” language of Hagood means the courts should avoid substantively analyzing a public disclosure issue. 816 F.3d at 577-78 (stressing the need to analyze the issue “with some specificity”).

Even assuming the district court committed legal error by failing to analyze this issue in enough detail, any error would be harmless. We are satisfied that enough of the fraud allegations, or alternatively, enough of the underlying facts, were publicly disclosed such that the public disclosure bar should have been triggered.

B

Since the public disclosure bar was triggered, did Amphastar qualify as an original source?

1

The first prong of the original source analysis asks whether Amphastar can show it had “knowledge that is both direct and independent.” Devlin, 84 F.3d at 361 n.5. To prove “direct” knowledge, Amphastar “must show that [it] had firsthand knowledge of the alleged fraud, and that [it] obtained this knowledge through [its] own labor unmediated by anything else.” Alcan, 197 F.3d at 1020. To prove “independent” knowledge, relators have to show they had relevant “evidence of fraud prior to the public disclosure of the allegations.” Devlin, 84 F.3d at 361 n.5.

Amphastar claimed it learned that Example 6 was false through its Pre-Litigation Experiments (direct) and knew this information before the Aventis I disclosures (independent). Amphastar Pharm., Inc. v. Aventis Pharma S.A., No. EDCV-09-0023, 2015 WL 4511573 , at *6 (C.D. Cal. July 20, 2015) (“Aventis VI”). Specifically, Amphastar argued that its Pre-Litigation Experiments showed that it could create “a bioequivalent enoxaparin with the same molecular weight and half-life properties as” the 618 Product by just following EP144. Id. Thus, Example 6, which showed a difference in half-life between the two products, must be false, it argued. Id.

2

So the crucial question is whether the district court’s factual findings derived from the July 2014 evidentiary hearing are clearly erroneous. The district court did a thorough job developing the record—eight witnesses (including four experts) testified, 300 exhibits were submitted, and the transcript for the hearing exceeds 1,100 pages. The district court found that Amphastar had no knowledge Example 6 was false before this fact was publicly disclosed in the Aventis I litigation. Aventis VI, 2015 WL 4511573 , at *6. The court concluded that Zhang’s testimony, that he knew Example 6 of the 618 Patent application was false before the Aventis I litigation began, was not credible. Id. It also concluded that Amphastar had no direct or independent knowledge that Aventis had made false reports to the FDA about its manufacturing process. Id. at *14 (“Indeed, in argument, Amphastar acknowledged that its only witness regarding its supposed ‘direct and independent knowledge,’ Zhang, did not offer such evidence.”).

The district court rejected Zhang’s testimony that he believed Example 6 was false before Aventis I for several reasons. For example, there was no written documentation of this seemingly important discovery. The only proof Zhang offered was a single statement in his deposition that he allegedly told two other people in passing, yet neither of these people testified to confirm these alleged conversations. Id. at *6 . The court also concluded that the Pre-Litigation Experiments did not actually produce enoxaparin by following EP144, and thus did not demonstrate that Example 6 was false—despite Zhang claiming this is how he first came to believe it. Id. at *8 .

Additional evidence supporting the district court’s conclusion was the notebooks documenting the Pre-Litigation Experiments; they provided critical evidence that Amphastar ended up relying on the 618 Patent to create a generic enoxaparin. Id. at *-9-14. This greatly undermined Am- phastar’s argument that it was able to duplicate the 618 Product using just EP144 disclosures and thus proved to itself that something was wrong with the 618 Patent. Amphastar even conceded that it is “not challenging the court’s factual findings regarding its pre-litigation experiments.”

The district court was also deeply troubled (for good reason) by Amphastar’s inconsistent prior statements about when it discovered the Example 6 error. Amphas-tar did not raise the Example 6 issue as a defense until a year after the Aventis I litigation began. Aventis VI, 2015 WL 4511573 , at *7. When it did raise the defense, in its June 7, 2004 motion, it recited, “based on facts recently developed in this litigation, Amphastar has added new affirmative defenses [and several counterclaims].” Id. (emphasis added by district court). Amphastar did not list the supposedly important Pre-Litigation Experiments among the evidence in support of its June 2004 contentions. Id. Also, the pre-suit disclosure letter written by Amphastar in 2008 and sent to the government before Aventis VI stated that “[djuring Amphas-tar’s litigation against Aventis, Amphas-tar discovered that Aventis has committed frauds against [the United States], which have resulted in false claims.” Id. at *8 (emphasis added by district court).

All of Amphastar’s legal arguments regarding direct and independent knowledge fail if the factual findings of the evidentia-ry hearing stand. Amphastar argues the district court incorrectly required Amphas-tar to show it had knowledge of Example 6’s falsity and of Aventis’s intent to deceive. Furthermore, the district court does use the phrases “false statement” and “knowledge of the alleged fraud” interchangeably, which is problematic. See, e.g., id. at *6. Amphastar only needed to prove it had direct and independent knowledge that Example 6 was false to be an original source—the district court should have spoken with more precision. Nevertheless, the factual conclusion that Amphastar had no knowledge of the Example 6 issue before it was disclosed makes this argument moot. Any error is harmless if Amphastar fails to overturn the factual findings.

Amphastar was required to offer compelling proof the district court erred to overcome the deferential clear error standard that governs factual findings. Anderson v. City of Bessemer, 470 U.S. 564, 573 , 105 S.Ct. 1504 , 84 L.Ed.2d 518 (1985) (describing the “clearly erroneous” standard). Yet, Amphastar does not even directly challenge the. district court’s factual findings, instead arguing that the district court did not adequately consider its March 17, 2003, experiment in considering whether Amphastar knew about the Example 6 issue through its own labor pre- Aventis I. However, the district court clearly concluded that no Pre-Litigation Experiment provided Amphastar this information. It discussed all of the issues with the Pre-Litigation Experiments in general and then provided specific examples. The district court is not required to list all one hundred plus' experiments and explain why each one did not provide this information, but instead it is presumed to have considered all the evidence unless something indicates otherwise. See W. Pac. Fisheries, Inc. v. SS President Grant, 730 F.2d 1280 , 1285 (9th Cir. 1984). Here, the district court considered “the evidence” and reached the not clearly erroneous conclusion that none of the Pre-Litigation Experiments produced enoxaparin utilizing just EP144 disclosures. Aventis VI, 2015 WL 4511573 , at *1, 8.

3

Amphastar further argues that the district court abused its discretion by allowing Aventis to present an argument at the evidentiary hearing that it copied 618 Patent processes in its Pre-Litigation Experiments. Amphastar served interrogatories to Aventis before the evidentiary hearing to disclose facts, witnesses, and other items supporting its arguments. Aventis did not disclose that it would make an argument that Amphastar copied the 618 Patent in response to the interrogatories or in its pre-evidentiary hearing brief. Amphastar believes this theory, and any related contentions, should have been excluded under Federal Rule of Civil Procedure 37(c)(1), which states: “If a party fails to provide information or identify a witness as required by Rule 26(a) or (e), the party is not allowed to use that information ... unless the failure was substantially justified or is harmless.”

However, Amphastar waived this argument by failing to raise it adequately below. Nowhere before the district court did Amphastar make a Rule 37(c)(1) objection or ask that evidence relating to the copying theory be excluded. At the evidentiary hearing’s closing arguments Amphastar did mention being caught off-guard by the “copying” theory, but got several facts wrong and never attempted to exclude the theory or raise a Rule 37 objection. Am-phastar did make a cursory Rule 37(b)(2)(A)(ii) request for sanctions and an objection in its post-hearing brief, alleging that “Aventis ... failed to provide a full and complete disclosure of its contentions.” But no citation to Rule 37(c)(1) was made.

Assuming arguendo that this issue was not waived and the district court did err, any error was harmless. Even if the “copying” theory were excluded, the record would still contain all of the other factual findings that indicate Amphastar lacked direct and independent knowledge. The inconsistencies, the notebooks, Zhang’s lack of credibility—all would remain. Amphastar would still bear the burden of prevailing on the jurisdictional bar issue. Alcan, 197 F.3d at 1018. It has not come close to doing so.

Ill

In its cross-appeal, Aventis contends that the district court erred in concluding it had no power to award attorneys’ fees to Aventis, and we agree. The relevant fee-shifting provision states:

31 U.S.C. § 3730 (d)(4) (emphasis added).

A

The district court relied on Branson v. Nott, 62 F.3d 287, 293 (9th Cir. 1995), to conclude that as a matter of law attorneys’ fees cannot be awarded under a fee-shifting statute if the underlying claim was dismissed for lack of subject matter juris diction. Aventis VI, 2015 WL 4511573 , at *16.

Branson and its progeny set out a two-part test for analyzing whether a district court can award attorneys’ fees when the underlying action is dismissed for lack of subject matter jurisdiction in a case such as this: (1) Does the fee-shifting provision contain an independent grant of subject matter jurisdiction? (2) If so, did the winning party “prevail?” 62 F.3d at 293 .

First, we look to see if the fee-shifting provision “provide[s] the district court with jurisdiction to grant an attorney fee award where subject matter jurisdiction to hear the underlying ... claim is lacking.” Id. Sometimes the answer is no. See, e.g., Zambrano v. I.N.S., 282 F.3d 1145 , 1150 (9th Cir. 2002) (concluding that the fee-shifting statute “does not provide an independent grant of jurisdiction” and therefore the district court lacks the jurisdiction to award attorneys’ fees). Sometimes the answer is yes. See, e.g., Alaska Right to Life Political Action Comm. v. Feldman, 504 F.3d 840, 852 (9th Cir. 2007) (holding that we “may award attorneys’ fees and costs even after dismissing for lack of jurisdiction”). And sometimes the answer is yes, but only with regard to the jurisdictional issues. Latch v. United States, 842 F.2d 1031 , 1033 n.4 (9th Cir. 1988) (per curiam) (explaining that 26 U.S.C. § 7430 allows fees to be awarded to a defendant that “prevailed on the only issue over which the district court properly had jurisdiction, i.e., the determination that it had no jurisdiction”).

Second, “even if the district court had the jurisdiction to impose ... attorney’s fees [under the fee-shifting provision], attorney’s fees are only available under that provision to a party who has ‘prevailed’ on the merits.” Branson, 62 F.3d at 293 . The district court skipped to the second-prong and reasonably concluded that Branson controlled, Aventis did not “prevail,” and it could not award attorneys’ fees. Aventis VI, 2015 WL 4511573 , at *16; see also Branson, 62 F.3d at 293 (“Where, as here, dismissal is mandated by a lack of subject matter jurisdiction, a defendant is not a ‘prevailing’ party within the meaning of § 1988.”).

B

But, Branson’s analysis of the second-prong has been called into question. In CRST Van Expedited Inc. v. E.E.O.C. the Supreme Court explained that “a favorable ruling on the merits is not a necessary predicate to find that a defendant has prevailed.” — U.S. -, 136 S.Ct. 1642, 1646 , 194 L.Ed.2d 707 (2016). It reversed an Eighth Circuit decision that declined to award attorneys’ fees because the defendant did not win “on the merits.” Id. at 1650, 1654 . The court elaborated at length about why defendants who prevail for various non-meritorious reasons should still be deemed “prevailing parties].” Id. at 1651-54 . In doing so it included an example where defendants prevailed for jurisdictional reasons, id. at 1653 , and never indicated that jurisdictional and non-jurisdictional victories should be treated differently. Finally, it made clear that its reasoning applied to other fee-shifting statutes, stating that “Congress has included the term ‘prevailing party’ in various fee-shifting statutes, and it has been the Court’s approach to interpret the term in a consistent manner.” Id. at 1646 . For these reasons, we conclude that the Supreme Court has effectively overruled Branson’s holding that when a defendant wins because the action is dismissed for lack of subject matter jurisdiction he is never a prevailing party. See Miller v. Gammie, 335 F.3d 889, 899-900 (9th Cir. 2003) (en banc) (explaining that a three-judge panel may no longer be bound by prior circuit precedent if it is “effectively overruled by subsequent Supreme Court decisions”).

With Branson out of the way, the question of whether Aventis is a prevailing party is an easy one: Yes. Aventis has spent eight years, quite a bit of money, and quite a bit of energy, fighting this lawsuit. This lawsuit has probably lasted longer than the vast majority of lawsuits that are resolved on the “merits.” If Api-phastar brought a frivolous action, the statutory scheme strongly indicates that this is the kind of case for which fee awards should be available to deter future frivolous plaintiffs. See CRST, 136 S.Ct. at 1652 . Common sense says that Aventis has won a significant victory and permanently changed the “legal relationship of the parties.” Id. at 1646 .

C

Amphastar challenges whether § 3730(d)(4) contains an independent grant of subject matter jurisdiction to award attorneys’ fees. We have located independent grants in some cases but not in others, as discussed above, but whether § 3730(d)(4) contains an independent grant of jurisdiction is a matter of first impression for this circuit. When deciding this question we have looked at the text and structure of the statute. See Latch, 842 F.2d at 1033 . We have also been more willing to allow a district court to award attorneys’ fees when the underlying issues concerned whether the district court had jurisdiction—since a court always has jurisdiction to determine its own jurisdiction. Id. at 1033 n.4.

Two courts in other circuits have addressed this question. Both conclude that § 3730(d)(4) provides an independent grant of jurisdiction to award attorneys’ fees even if the underlying action is dismissed for lack of jurisdiction. See United States ex rel. Grynberg v. Praxair, Inc., 389 F.3d 1038, 1055-58 (10th Cir. 2004); United States ex rel. Atkinson v. Pa. Shipbuilding Co., 528 F.Supp.2d 533, 540 (E.D. Pa. 2007). Both relied on the text and structure of the statute. Id. Atkinson observed that “[eliminating jurisdiction to award attorney fees and expenses in such cases would have the illogical consequence of disallowing fees in cases specifically identified by Congress as undesirable.” 528 F.Supp.2d at 539-40 .

The False Claims Act jurisdictional bar covers a huge number of cases. To rule that a district court cannot award attorneys’ fees even when it determines that a Relator brought a frivolous suit just because the jurisdictional bar applies would undermine one of the key purposes of the 1986 amendments to the False Claims Act—“to discourage ‘parasitic’ suits brought by individuals with no information of their own to contribute to the suit.” United States v. Johnson Controls, Inc., 457 F.3d 1009, 1017 (9th Cir. 2006) (quoting Seal 1 v. Seal A, 255 F.3d 1154, 1158 (9th Cir. 2001)). It is consistent with the statutory scheme that Aventis can receive attorneys’ fees from Amphastar if its claim was frivolous, given the immense amount of resources and time this action has cost everyone. Cf. CRST, 136 S.Ct. at 1652-53 (discussing how in the context of Title VII of the Civil Rights Act, “significant attorney time and expenditure may have gone into contesting the claim [in cases like these]” and “Congress could not have intended to bar defendants from obtaining attorney’s fees in these cases”).

Given Latch , the text and structure of the statute, and Grynberg and Atkinson, we conclude that § 3730(d)(4) contains an independent grant of subject matter jurisdiction at least to the extent that a court is able to award attorneys’ fees even where the issue is whether the court had subject matter jurisdiction at all.

D

We are persuaded that the district court had subject matter jurisdiction over the attorneys’ fees issue and that Aventis is the prevailing party. As such, the district court may award attorneys’ fees to Aventis if it determines that Amphastar’s claim “was clearly frivolous, clearly vexatious, or brought primarily for purposes of harassment.” § 3730(d)(4).

IV

Amphastar’s allegations in this action were based on publicly disclosed information, and it lacked the direct and independent knowledge needed to be an original source. We uphold the judgment of the district court in favor of Aventis on the merits. However, because the False Claims Act’s fee-shifting provision contains an independent grant of subject matter jurisdiction and because a party who wins a lawsuit on a non-merits issue is a “prevailing party,” the conclusion of the district court that it could not award attorneys’ fees is erroneous, and the case is remanded for resolution of the attorneys’ fees issue.

Therefore, the judgment of the district court in No. 15-56122 is AFFIRMED, the appeal in No. 14-56382 is DISMISSED, the judgment in No. 15-56209 is REVERSED, and the case is REMANDED for further proceedings consistent with this opinion.

. A qui tam action is one in which a private party (the relator) brings a lawsuit on behalf of the United States government alleging fraud in return for a portion of any- damages awarded. See 31 U.S.C. § 3730 (b). . Aventis Pharma S.A. and Aventis Pharmaceuticals, Inc. (the American subsidiary) merged with and into Sanofi-Aventis S.A. during this litigation. The defendants-appellants are collectively referred to as Aventis. . Referred to as the "618 Product” by previous courts because it was eventually covered by Patent Number 5,389,618. . Example 6 of the patent application describes data from studies that purport to show that the 618 Product has a significantly longer half-life than the EP144 Product, enabling the same effect to be achieved with lower dosages. . This patent was reissued as U.S. Reissue Patent No. 38,743 in 2005. Both patents are collectively referred to herein as the "618 Patent.” . The initial 2003 patent infringement suit by Aventis led to the finding, on summary judgment, that Amphastar’s affirmative defense on inequitable conduct grounds succeeded because (1) Aventis’s false representations and nondisclosures to the USPTO were material, and (2) it had intent to deceive the USPTO. Aventis Pharma S.A. v. Amphastar Pharm., Inc., 390 F.Supp.2d 952 (C.D. Cal. 2005) {“Aventis I"). The Federal Circuit affirmed in part but reversed in part. 176 Fed.Appx. 117 (Fed. Cir. 2006) ("Aventis IT’) (affirming the material false representations and nondisclo-sures finding, but vacating the intent to deceive conclusion and remanding the case to the district court for a trial on the intent to deceive issue). On remand, the district court again found that Aventis did intend to deceive the USPTO; this time the decision was affirmed on appeal. Aventis Pharma S.A. v. Amphastar Pharm., Inc., 475 F.Supp.2d 970, 994 (C.D. Cal. 2007) (“Aventis III"), aff'd, 525 F.3d 1334 (Fed. Cir. 2008) ("Aventis IV"). . Said action was filed pursuant to 31 U.S.C. § 3730 , which was amended on March 23, 2010, but the amendments are not retroactive. Graham Cnty. Soil & Water Conservation Dist. v. United States ex rel. Wilson, 559 U.S. 280 , 283 n.1, 130 S.Ct. 1396 , 176 L.Ed.2d. 225 (2010). We analyze the statute as it existed when Amphastar first filed its complaint in 2009. . Direct knowledge is firsthand knowledge of the alleged fraud obtained by the relator’s own labor. United States v. Alcan Elec. & Eng’g, Inc., 197 F.3d 1014 , 1020 (9th Cir. 1999). Independent knowledge is knowledge obtained by the relator before the evidence of the fraud was disclosed to the public. United States ex rel. Devlin v. California, 84 F.3d 358 , 361 n.5 (9th Cir. 1996). . Whether subject matter jurisdiction exists "is a question of law reviewed de novo.” United States ex rel. Biddle v. Bd. of Trs. of Leland Stanford, Jr. Univ., 161 F.3d 533, 535 (9th Cir. 1998). "The district court’s findings of fact relevant to its determination of subject matter jurisdiction are reviewed for clear error.” Id. District court evidentiary rulings are reviewed for abuse of discretion. Gribben v. United Parcel Serv., Inc., 528 F.3d 1166, 1171 (9th Cir. 2008). Other legal issues, such as the attorneys’ fees issue, are reviewed de novo. See Salve Regina Coll. v. Russell, 499 U.S. 225, 231 , 111 S.Ct. 1217 , 113 L.Ed.2d 190 (1991). . These include the Amended Answer and Counterclaim filed by Amphastar and the Motion for Leave (After the Fact) to File its Amended Answer and Counterclaim (Amend Motion) in Aventis I, which added new affirmative defenses and several antitrust claims "based on facts recently developed in this litigation.” . Amphastar also argues that it gained direct and independent knowledge that Aventis com mitted fraud during the Aventis I discovery process. However, this argument is not relevant to the original source analysis. Even if Amphastar did gain knowledge of the fraud because of the Aventis I discovery process, a relator’s involvement in prior litigation that constituted a public disclosure does not make it an original source when it learned of the alleged fraud solely through discovery of materials that Could potentially be accessed by the public. See, e.g., United States ex rel. Kreindler & Kreindler v. United Techs. Corp., 985 F.2d 1148, 1158-59 (2d Cir. 1993); United States ex rel. Stinson, Lyons, Gerlin & Bustamante, P.A. v. Prudential Ins. Co., 944 F.2d 1149, 1160-61 (3d Cir. 1991). To qualify as an original source Amphastar must show it gained some direct and independent knowledge from its Pre-Litigation Experiments, since this is the only non-Aventis I discovery source of information from which it claims it gained direct and independent knowledge. This is presumably why Amphastar devoted almost all of its efforts at the July 2014 evi-dentiary hearing to proving that it learned Example 6 was false because of its Pre-Litigation Experiments. . Amphastar originally produced copies of the notebooks documenting the Pre-Litigation Experiments that were missing certain pages. Id. at *9. However, Amphastar actually knew that the original notebooks contained some of those pages and did not inform the court of this discrepancy until it was caught. Id. at *9-11. . Because we conclude that Amphastar is not an original source since it lacks direct and independent knowledge of some of the information upon which the qui tam allegations are based, we need not address, and choose not to consider, Aventis's interlocutory appeal (No. 14-56382). That appeal raised the issue of whether the district court erred by concluding that Amphastar's pre-suit disclosure letter satisfied the pre-suit disclosure requirement of § 3730(e)(4)(B). In light of our conclusions above, the interlocutory appeal will be dismissed as moot. See Earth Island Inst. v. U.S. Forest Serv., 442 F.3d 1147, 1157 (9th Cir. 2006). . The Supreme Court tells us that we should look to whether there was a “material alteration of the legal relationship of the parties.” Id. at 1646. We should also look to whether deeming a party “prevailing” and awarding fees for frivolous litigation furthers the purpose of the statutory scheme. Id. at 1652. Given that the purpose of fee-shifting statutes is to deter frivolous litigation, and since fees can only be awarded if the plaintiff's action is deemed frivolous, the answer is almost always yes. See id. (“Imposing an on-the-merits requirement for a defendant to obtain prevailing party status would undermine that congressional policy by blocking a whole category of defendants for whom Congress wished to make fee awards available.”). . Congress enacted two related provisions to further that goal in 1986: the public disclosure bar (§ 3730(e)(4)) and the attorneys’ fees provision for prevailing defendants (§ 3730(d)(4)).