TINDER, Circuit Judge.
The collapse of investment manager Sentinel Management Group, Inc., in the summer of 2007 left its customers in a lurch. Instead of maintaining customer assets in segregated accounts as required by law, Sentinel had pledged hundreds of millions of dollars in customer assets to secure an overnight loan at the Bank of New York, now Bank of New York Mellon. This left the Bank in a secured position on Sentinel’s $312 million loan but its customers out millions. Once Sentinel filed for bankruptcy, Sentinel’s Liquidation Trustee, Frederick J. Grede, brought a variety of claims against the Bank—including fraudulent transfer, equitable subordination, and illegal contract—to dislodge the Bank’s secured position. After extensive proceedings, including a seventeen-day bench trial, the district court rejected all of the Trustee’s claims. Although we appreciate the district court’s painstaking efforts, we cannot agree with its conclusion that Sentinel’s failure to keep client funds properly segregated was insufficient to show an actual intent to hinder, delay, or defraud. We also find significant inconsistencies in both the factual and legal findings of the district court with respect to the equitable subordination claim. For these reasons, we reverse the judgment of the district court with respect to Grede’s fraudulent transfer and equitable subordination claims.
I. Factual Background
Even though we find some inconsistencies in the thirty-nine-page opinion of the district court, its comprehensive review of the evidence still provides a useful starting point for our discussion. See Grede v. Bank of N.Y. Mellon, 441 B.R. 864 (N.D.Ill.2010). The district court’s findings of fact, of course, “are. entitled to great deference and shall not be set aside unless they are clearly erroneous.” Gaffney v. Riverboat Servs. of Ind., Inc., 451 F.3d 424, 447 (7th Cir.2006). Nonetheless, we review the district court’s findings of law—including the district court’s determination of actual intent to hinder, delay, or defraud—de novo. Johnson v. West, 218 F.3d 725, 729 (7th Cir.2000).
Before filing for bankruptcy in August 2007, Sentinel was an investment manager that marketed itself to its customers as providing a safe place to put their excess capital, assuring solid short-term returns, but also promising ready access to the capital. Sentinel’s customers were not typical investors; most of them were futures commission merchants (FCMs), which operate in the commodity industry akin to the securities industry’s broker-dealers. In Sentinel’s hands, FCMs’ client money could, in compliance with industry regulations governing such funds, earn a decent return while maintaining the liquidity FCMs need. “Sentinel has constructed a fail-safe system that virtually eliminates risk from short term investing,” proclaimed Sentinel’s website in 2004.
To accept capital from its FCM customers, Sentinel had to register as a FCM, but it did not solicit or accept orders for futures contracts. Sentinel received a no-action” letter from the Commodity Futures Trading Commission (CFTC) exempting it from certain requirements applicable to FCMs. But Sen[663]*663tinel represented that it would maintain customer funds in segregated accounts - as required under the Commodity Exchange Act, 7 U.S.C. § 1 et seq. Maintaining segregation meant that at all times a- customer’s accounts held assets equal to the amount Sentinel owed the customer, and that Sentinel treated and dealt with the assets “as belonging to such customer.” 7 U.S.C. § 6d(a)(2) (“Such money, securities, and property shall be separately accounted for and shall not be commingled with the funds of such commission merchant or be used to margin or guarantee the trades or contracts, or to secure or extend the credit, of any customer or person other than the one for whom the same are held ... ”).
Maintaining segregation serves as commodity customers’ primary legal protection against wrongdoing or insolvency by FCMs and their depositories, similar to depositors’ Federal Deposit Insurance Corporation protection, see 12 U.S.C. § 1811 et seq., or securities investors’ Securities Investor Protection Corporation protection, see 15 U.S.C. § 78aaa et seq. Sentinel also served other investors such as hedge funds and commodity pools, and as early as 2005, began maintaining a house account for its own trading activity to benefit Sentinel insiders. In 2006, Sentinel represented that non-FCM entities made up about one-third of its customer base. By 2007, Sentinel held about $1.5 billion in customer assets but maintained only $3 million or less in net capital.
Sentinel pooled customer assets in' various portfolios, depending on whether the customer assets were CFTC-regulateB- assets of FCMs or unregulated funds such as hedge funds or FCMs’ proprietary funds. But Sentinel handled “its and its customers’ assets as a single, undifferentiated pool of cash and securities.” Grede, 441 B.R. at 874. When customers wanted their capital back, Sentinel could sell securities or borrow the money. Sentinel’s borrowing practices, .and in particular an overnight loan it maintained with the Bank of New -York, is this appeal’s focal point: This arrangement allowed Sentinel to borrow large amounts of cash while pledging customers’ securities as collateral.
Sentinel’s relationship with the Bank began in 1997 in the Bank’s institutional-custody division but within months moved to the clearing division (technically dubbed broker-dealer services) because Sentinel actively traded securities and frequently financed transaction settlements. Under the old arrangement, for each segregated account, Sentinel had a cash account for customer deposits and withdrawals. Assets could not leave segregation without a corresponding transfer from a cash account. But the risks of overdrafts prompted a switch to an environment where securities would be bought and sold from clearing accounts lienable by the Bank. In an email, one bank official said in reference to Sentinel’s original arrangement that “THIS ACCOUNT IS AN ACCIDENT WAITING TO HAPPEN.... I AM NOTIFYING YOU THAT I NO LONGER FEEL COMFORTABLE CLEARING THESE TRANSACTIONS AND REQUEST AN IMMEDIATE RESPONSE FROM YOU. THANK YOU.”
Under the new arrangement, Sentinel maintained three types of accounts at the Bank. First, clearing accounts allowed Sentinel to buy or sell securities, including government, corporate, and foreign securities and securities traded with physical certificates. The Bank maintained the right to place a lien on the assets in clearing accounts. Second, Sentinel maintained an overnight loan account in conjunction with its secured line of credit. To borrow on the line of credit, Sentinel would call bank officials to confirm whether it had [664]*664sufficient assets in lienable accounts to serve as collateral. A senior bank executive had to approve requests that put the line of credit above a predetermined “guidance line.” Third, Sentinel maintained segregated accounts that held assets that could not be subject to any bank lien. These included accounts (corresponding with the lienable clearing accounts) for government, corporate, and foreign securities but no corresponding segregated account for physical securities. To receive FCM funds in the segregated accounts, the Bank countersigned letters acknowledging that the funds belonged to the customers and that the accounts would “not be subject to your lien or offset for, and on account of, any indebtedness now or hereafter owing us to you.... ” Thé agreement between Sentinel and the Bank provided that the “Bank will not have, and will not assert, any claim or lien against Securities held in a Segregated Account nor will Bank grant any third party ... any interest in such Securities.”
Sentinel could independently transfer assets between accounts by issuing electronic desegregation instructions without significant bank knowledge or involvement. This system allowed for hundreds of thousands of trades worth trillions of dollars every day at the Bank. Sentinel maintained responsibility for keeping assets at appropriate levels of segregation. The Bank’s main concern was ensuring Sentinel had sufficient collateral in the lienable accounts to keep its overnight loan secured. In fact, at no point does it appear that the Bank was under-secured. If Sentinel sought to extend the line of credit beyond the value of the assets held in the lienable accounts, the Bank made sure Sentinel moved enough collateral into the lienable accounts. Sentinel used cash from the overnight loan for customer re-demptions or failed trades and provided collateral in the form of the customers’ redeemed securities. When customers redeemed investments, Sentinel could provide cash, via the loan, without waiting for the securities to sell. This arrangement did not violate segregation requirements. When a customer cashed out, the amount needed in segregation dropped by the amount lent by the Bank via the line of credit. The line of credit was in turn secured by assets moved out of customers’ segregated accounts and into clearing accounts.
But in 2001, and increasingly in 2004, Sentinel started using the loan to fund its own proprietary repurchase arrangements with counterparties such as FIMAT USA and Cantor Fitzgerald & Co. Sentinel would finance most of a security’s purchase price by transferring ownership of the security to a counterparty, who would lend Sentinel an amount of cash equal to a percent of the asset’s market value. Sentinel used the overnight loan to cover the difference (known as a “haircut”) between the security’s cost and the repo loan. Sentinel had to buy the security back at some point for the amount loaned plus interest. By 2007, Sentinel held more than $2 billion in securities through repo arrangements. Meanwhile, Sentinel’s guidance line for the Bank loan grew from $30 million pre-May 2004, to $55 million in May 2004, to $95 million in December 2004, to $175 million in June 2005, to $300 million in September 2006. The average loan balance from June 1, 2007, to August 13, 2007; was $369 million. The line topped out at $573 million at one point, while all along customer assets served as collateral. In 2004, Sentinel faced a segregation shortfall of about $150 million, and by July 2007, that figure reached nearly $1 billion.
During the summer of 2007, the cloud of a liquidity and credit crunch settled in. Repurchase lenders became nervous. The type of securities Sentinel held became a [665]*665focus of the market as counterparties stopped accepting securities previously used as collateral. They wanted cash. But the crunch prevented selling-the securities. Cash was tough to obtain. As Sentinel turned increasingly to its line of credit for cash, the Bank’s thirst for the highest-rated, most-liquid securities to secure the loan intensified.
On June 1, a counterparty returned $100 million in physical securities, and as a result, the Bank loan jumped from $259.7 million to $353 million over the course of a day. To meet the Bank’s demands for collateral, Sentinel moved about $88 million in government securities from segregated accounts to the lienable account. There was no way to maintain segregation levels via the returned physical securities because Sentinel did not keep segregated accounts for physical securities. Sentinel’s segregation deficit grew to $644 million. On June 13, the Bank became suspicious, and a managing director emailed various bank officials involved with the Sentinel account, asking how Sentinel had “so much collateral? With less than $20MM in capital I have to assume most of this collateral is for somebody else’s benefit. Do we really have rights on the whole $300MM? ?”1 After speaking to several bank officers, a Bank of New York client executive responded, “We have a clearing agreement which gives us a full lien on the box position outlined below.” The client executive testified that this was a well-advised and carefully worded statement, but both the managing director and the client executive knew Sentinel had an agreement that gave the Bank a lien on any securities in clearing accounts. Grede, 441 B.R. at 889-90. Then on June 26, a counterparty returned $166 million in physical securities. The Bank loan bal-’ anee accordingly grew to $497.5 million. For collateral, Sentinel moved $66.6 million in government securities out of segregated accounts and into the lienable account. These securities, however, were not enough for the Bank. In turn, Sentinel pledged $165 million in physical securities, and the segregation deficiency grew to $667 million. On June 27, Sentinel’s loan balanced peaked at $573 million. Two days later, the Bank told Sentinel it would no longer accept physical securities as -collateral. That day, Sentinel transferred $166 million in corporate securities from segregated accounts to the lienable account. Sentinel’s under-segregation problem grew to $813 million.
A similar transaction occurred on July 17, with a counterparty returning about $150 million in corporate securities. Sentinel transferred $84 million in corporate securities from a segregated account to a lienable account. The Bank loan settled at $496.9 million and Sentinel’s segregation shortfall grew to $935 million. At the month’s end, Sentinel briefly sent capital in the other direction. On July 30, Sentinel moved $248 million in corporate securities back into segregation from a lienable account and on July 31, $263 million in government securities back into segregation from a lienable account. Yet that same day, Sentinel moved $289 million in corporate securities from a segregated account to a lienable account. Sentinel’s loan settled at $356 million and its segregation deficit at $700 million.
After these transactions, Sentinel could not hang on and told customers on August 13 that it was halting redemptions because of problems in the credit markets. Once Sentinel told the Bank about this decision [666]*666the next day, the Bank cut Sentinel’s remote access to its systems, sent its officials to Sentinel’s offices, demanded full repayment of the loan, and threatened to- liquidate the collateral. Sentinel filed for bankruptcy on August 17, owing the Bank $312,247,000.
Plaintiff Frederick J. Grede was appointed Chapter 11 Trustee for Sentinel’s estate and, subsequent- to the Chapter 11 plan’s confirmation, the Trustee of the Sentinel Liquidation Trust. The Bank filed a $312 million claim as the only secured creditor. Grede filed an adversary proceeding against the Bank alleging that Sentinel fraudulently used customer assets to' finance the loan to cover its house trading activity. Grede further alleged that the Bank knew about it and, as a result, acted inequitably and unlawfully. Grede brought claims of fraudulent transfer under the Bankruptcy Code and state law, 11 U.S.C. §§ 544(b)(1), 548(a)(1)(A); 740 ILES 160/5(a)(l), and preferential transfer, 11 U.S.C. § 547(b), all to avoid the Bank’s lien, see 11 U.S.C. § 550(a). Grede also brought claims of equitable subordinar tion of the Bank’s claim, 11 U.S.C. § 510(c), and invalidation of the Bank’s lien, 11 U.S.C. § 506(d), among others. The district court dismissed the lien invalidation count on the pleadings, Grede v. Bank of New York, No. 08 C 2582, 2009 WL 188460, at *8 (N.D.Ill. Jan. 27, 2009), and the Bank moved for summary judgment on the other claims. The court reserved ruling on the Bank’s motion and held a bench trial that lasted seventeen days. After hearing from more than a dozen witnesses, listening to audio recordings between bank and Sentinel officials, and reviewing hundreds of exhibits, the district court ruled in the Bank’s favor on the remaining counts. The court found that Grede had “failed to prove that Sentinel made the Transfers with the actual intent to hinder, delay or defraud its creditors.” Grede, 441 B.R. at 881. The court also rejected the preference claim because the Bank was over-collateralized on the transfer dates. Id. at 886. With respect to equitable subordination, the court rejected Grede’s claim because it did not believe that the Bank’s conduct was “egregious or conscience shocking.” Id. at 901. Moreover, the court found that the Bank employees “had no legal obligation ... to seek out or analyze the data” that would have revealed Sentinel’s misuse of the segregated funds. Id. at 895.
II. Analysis
In the district court, Grede advanced three arguments why the Bank of New York should be dislodged from its secured position. First, Grede argued that Sentinel acted with actual intent to hinder, delay, or defraud when it borrowed money from the Bank, and thus, the Bank’s lien should be avoided. Second, Grede argued that the Bank engaged in inequitable conduct when it allowed Sentinel to borrow money, and as a result, the Bank’s lien should be subordinated to the claims of unsecured creditors. Third, Grede argued that Sentinel’s contracts with the Bank violated the law on their face, so the Bank’s lien should be invalidated. We address each of Grede’s arguments in turn.
A. Fraudulent Transfer
11 U.S.C. § 548(a)(1)(A) allows the avoidance of any transfer of an interest in the debtor’s property if the debtor made the transfer “with actual intent to hinder, delay, or defraud” another creditor. Grede claims that the transfers of customer assets out of segregation and into the liena-ble accounts (which Sentinel used as collateral for its overnight loan with the Bank of New York) in June and July 2007 constituted fraudulent transfers under 11 U.S.C. [667]*667§§ 548(a)(1)(A) & 544(b), and should thus be avoided.
At the conclusion of the bench trial, the district court acknowledged that Sentinel “[was already insolvent at the time of the transfers” and had “missed] creditor assets.” But the district court did not believe such behavior was enough to prove that Sentinel possessed the actual intent to hinder, delay, or defraud other creditors besides the Bank (including its FCM clients), as required to avoid a lien under 11 U.S.C. § 548(a)(1)(A). In reaching this conclusion, the district court relied on Gre-de’s expert witness, James Feldman, who had testified that “three of the transfers in question ‘had to do with the closing out of repo positions[,]’ and the remaining two were related to what Feldman called ‘structuring of collateral, the movement of securities between accounts.’ ” Based on this testimony, the district judge appeared to believe that Sentinel had robbed Peter (Sentinel’s FCM clients) to pay Paul (the Bank of New York) in the months before it filed for Chapter 11 bankruptcy. While the district court’s opinion certainly did not condone such behavior, it concluded that this behavior was not enough to show that Sentinel had actual intent to hinder, delay, or defraud its FCM clients. Rather, the opinion characterized Sentinel’s behavior as a desperate “attempt to stay in business.”
This finding that Sentinel’s pledge of segregated funds as collateral for loans with the Bank of New York was driven by a desire to stay in business correctly identified the motive. Nonetheless, we disagree with the district court’s legal conclusion that such motivation was insufficient to constitute actual intent to hinder, delay, or defraud Sentinel’s FCM clients. Such a result too narrowly construes the concept of actual intent to hinder, delay, or defraud. When Sentinel pledged the funds that were supposed to remain segregated for its FCM clients, Sentinel’s primary purpose may not have been to render the funds permanently unavailable to these clients (although Sentinel falsely reported to both its FCM clients and the CFTC that the funds remained in segregation). But Sentinel certainly should have seen this result as a natural consequence of its actions. In our legal system, “every person is presumed to intend the natural consequences of his acts.” In re Danville Hotel Co., 38 F.2d 10, 21 (7th Cir.1930); see also Reno v. Bossier Parish Sch. Bd., 520 U.S. 471, 487, 117 S.Ct. 1491, 137 L.Ed.2d 730 (1997); Trzcinski v. Am. Cas. Co., 953 F.2d 307, 313 (7th Cir.1992).
Consequently, we conclude that Sentinel’s transfers of segregated funds into its clearing accounts demonstrate an “actual intent to hinder, delay, or defraud” under 11 U.S.C. § 548(a)(1)(A). To treat these transfers as fraudulent is consistent with our construction of actual intent to defraud in other contexts. For example, in United States v. Segal, 644 F.3d 364, 367 (7th Cir.2011), Michael Segal appealed his conviction of, among other crimes, mail and wire fraud under 18 U.S.C. §§ 1341, 1343. Segal argued that his fraud convictions should be overturned because he lacked “a specific intent to cause injury.” Id. We rejected this argument, finding that a defendant could have an actual intent to defraud without having an actual intent to cause harm. Id.
Similarly, in United States v. Davuluri, 239 F.3d 902, 906 (7th Cir.2001), the defendant, Surya Davuluri, also appealed his conviction of mail and wire fraud under 18 U.S.C. §§ 1341, 1343. Davuluri argued that the undisputed evidence demonstrated that “he always intended only to earn profits for [his victim], rather than imposing any kind of loss on him,” and that such intentions were inconsistent with an actual [668]*668intent to defraud. Id. We rejected Davu-luri’s argument. Even if Davuluri had possessed nothing but the best intentions for his victim, we held that “a rational jury could find beyond a reasonable doubt that [Davaluri had] imposed a large risk of loss on [his victim]. Exposing the victim to a substantial risk of loss of which the victim is unaware can satisfy the intent requirement.” Id.; cf. United States v. Hamilton, 499 F.3d 734, 736 (7th Cir.2007) (Fraud is “not excused just because you had an honest intention of replacing the money.”).
Like Davuluri, Sentinel exposed its FCM clients to a substantial risk of loss of which they were unaware when it pledged funds that were supposed to remain segregated for the FCM clients as collateral for Sentinel’s overnight loans with the Bank of New York. Even though the district court found that Sentinel’s pledge was not an attempt “to drain its assets and make them unavailable to other creditors,” we held in Davuluri that someone who has the best intentions can still possess an actual intent to defraud. 239 F.3d at 906. Consequently, even if we assume that Sentinel had the best intentions for its FCM clients when it pledged the segregated funds, the fact remains that Sentinel knowingly exposed its FCM clients to a substantial risk of loss of which they were unaware.
Sentinel’s pledge of the segregated funds as collateral for its own loan becomes particularly egregious when viewed in light of the legal requirements imposed on Sentinel by the Commodity Exchange Act (CEA). Again, even if we assume that Sentinel eventually intended to replace the segregated funds and earn greater returns for their FCM clients, Sentinel knew that its pledge of the segregated funds violated the CEA. The CEA exists explicitly for the purpose of “ensuring] the financial integrity of all transactions” involving FCMs, “avoid[ing] systemic risk,” and “protecting] all market participants from ... misuses of customer assets.” 7 U.S.C. § 5(b). In order to further these aims, the CEA requires that the “money, securities, and property [belonging to clients] shall be separately accounted for and shall not be commingled with the funds of such commission merchant.” 7 U.S.C. § 6d(a)(2). Moreover, 7 U.S.C. § 6d(b) makes it “unlawful” for an FCM “to hold, dispose of, or use any such money, securities, or property as belonging to the depositing futures commission merchant.”
The language of the CEA makes clear that Sentinel did more than just expose its FCM clients to a substantial risk of loss of which they were unaware; Sentinel, in an unlawful manner, exposed its FCM clients to a substantial risk of loss of which they were unaware. Thus, even if Sentinel did not intend to harm its FCM clients, Sentinel’s intentions were hardly innocent. For this reason, we find that Sentinel’s actions, as determined by the factual findings of the district court, demonstrate an actual intent to hinder, delay, or defraud. As such, Grede should be able to avoid the Bank of New York’s lien under 11 U.S.C. § 548(a)(1)(A).2
[669]*669B. Equitable Subordination
Courts will subordinate a claim under 11 U.S.C. § 510(c) when the claimant creditor engaged in inequitable conduct that injured other creditors or conferred an unfair advantage on the claimant, but not when subordination is inconsistent with the Bankruptcy Code. See In re Kreisler, 546 F.3d 863, 866 (7th Cir.2008) (quoting United States v. Noland, 517 U.S. 535, 538-39, 116 S.Ct. 1524, 134 L.Ed.2d 748 (1996)). -'“Equitable subordination allows the bankruptcy court to reprioritize a claim if-it determines that the claimant is guilty of misconduct that injures other creditors or confers an unfair advantage on the claimant.” Id. (citing In re Lifschultz Fast Freight, 132 F.3d 339, 344 (7th Cir.1997)). Typically, the misconduct that courts have deemed sufficiently inequitable to merit this remedy has fallen within one of three areas: A “(1) fraud, illegality, breach of fiduciary duties; (2) under-capitalization; [or] (3) claimant’s use of the debtor as a mere instrumentality or alter ego.” Lifschultz, 132 F.3d at 345 (quoting In re Missionary Baptist Found, of Am., 712 F.2d 206, 212 (5th Cir.1983)). Keeping this general guidance in mind, we must decide here whether the Bank of New York’s conduct was sufficiently inequitable to merit the subordination of its lien.
“Equitable subordination means that a court has chosen to disregard an otherwise legally valid transaction.” Lifschultz, 132 F.3d at 347. In the past, our court—like other courts—has treaded very carefully before disregarding an otherwise legally valid transaction. See, e.g., Lifs-chultz, 132 F.3d at 347 (citing-At re Mobile Steel Co., 563 F.2d 692, 701-02 (5th Cir. 1977)). Two fundamental concerns have motivated our hesitance to invoke the doctrine of equitable subordination: (1) the upsetting of a claimant’s legitimate expectations, and (2) the spawning of legal uncertainty that courts will refuse to honor otherwise binding agreements “on ámor-phous grounds of equity,” increasing- everyone’s credit costs. Id.
Besides these two concerns, the difficulty of proving that a creditor has engaged in inequitable behavior has further increased our hesitance to apply the doctrine of equitable subordination. For example, the question of “ ‘whether a party has acted opportunistically,’ ” is quite subjective. Id. at 349 (quoting David A. Skeel, Jr., Markets, Courts, and the Brave New World of Bankruptcy Theory, 1993 Wis. L.Rev. 465, 506). There are simply no clear rules for determining whether underhanded behavior occurred. Id. (“Equitable subordination relies on courts’ peering behind the veil of formally' unimpeachable legal arrangements to detect the economic reality beneath.”). Underhanded behavior is typically clearest, however, when “corporate insiders [have attempted] to convert their equity interests into secured debt in anticipation of bankruptcy.” Kham & Nate’s Shoes No. 2, Inc. v. First Bank of Whiting, 908 F.2d 1351, 1356 (7th Cir. 1990). Consequently, courts have most frequently invoked this doctrine against corporate insiders, requiring them “to return to their position at the end of the line.” Id.
Proving that an outside creditor behaved inequitably in anticipation of the debtor’s [670]*670bankruptcy is much more difficult; the interests of an outside creditor are not necessarily aligned with the interests of the debtor (or with the interests of the debtor’s shareholders). As a result, courts have been particularly hesitant to invoke the doctrine of equitable subordination outside of cases involving insiders of closely held corporations. Lifschultz, 132 F.3d at 343. Some bankruptcy courts have even required wrongful conduct that rises to the level of “gross and egregious,” “tantamount to fraud, misrepresentation, overreaching .or spoliation,” or “involving moral turpitude” before equitably subordinating an outside creditor’s claim. In re Granite Partners, L.P., 210 B.R. 508, 515 (Bankr. S.D.N.Y.1997) (quotations and citation omitted). Consequently, it is not surprising that “[c]ases subordinating the claims of creditors that dealt at arm’s length with the debtor,” such as the case at hand, “are few and far between.” Kham, 908 F.2d at 1356.
In the past, our court has not directly addressed the degree of wrongful conduct sufficient to invoke the doctrine of equitable subordination against an outside creditor, so the district court here looked to decisions outside our circuit. Relying primarily on Granite Partners, 210 B.R. at 515, the district court determined that , equitable subordination would be inappropriate in this case unless the Bank’s behavior had been “egregious and conscience shocking.” Grede, 441 B.R. at 886, 901. The district court did not believe that the Bank’s behavior had sunk to this level, and thus, refused to subordinate the Bank’s claim.
But in reaching the conclusion that the Bank’s behavior was neither egregious nor conscience-shocking, the district court relied upon factual findings that were internally inconsistent. Although we normally give great deference to the district court’s factual findings after a full bench trial, we cannot extend the same deference to internally inconsistent factual findings, which are, by definition, clearly erroneous. See United States v. Sablotny, 21 F.3d 747, 751 (7th CirJ.994) (“The district court’s factual findings ... are not internally inconsistent, and are thus not in clear error.”); see also Anderson v. City of Bessemer, N.C., 470 U.S. 564, 575, 105 S.Ct. 1504, 84 L.Ed.2d 518 (1985) (“[W]hen a trial judge’s finding is based on his decision to credit the testimony of one of two or more witnesses ... that finding, if not internally inconsistent, can virtually never be clear error.”).
In particular, the district court appears to contradict itself regarding the extent of the Bank’s knowledge before Sentinel’s cbllapse. Approximately halfway through its opinion, the district court states, “[T]he evidence at trial revealed the Bank’s knowledge that Sentinel insiders were using at least some of the loan proceeds for their own purposes.” Grede, 441 B.R. at 883. This statement indicates that the Bank of New York knew Sentinel was engaging in wrongful conduct before its collapse. In contrast, the district court later states, “I do find credible, if not at all admirable, the testimony of the Bank employees that they neither knew nor turned a blind eye to the improper actions of Sentinel.” Id. at 894. If the Bank knew that Sentinel insiders were misusing the loan proceeds, then how could it be the case that bank employees “neither knew nor turned a blind eye to the improper actions of Sentinel”? The two factual findings, though in different parts of the opinion, make no sense when juxtaposed.
The fact that the Bank knew Sentinel insiders were misusing the loan proceeds before Sentinel’s collapse renders other statements in the district court’s opinion equally puzzling. For instance, at one [671]*671point in the opinion, the district court concludes:
If BNYM should have been more diligent with regard to verifying the source of collateral, such a lack of care does not rise to the level of the egregious misconduct necessary for equitable subordination. The fact remains that BNYM had little reason to conduct such a verification and could rely on representations and warranties. Notwithstanding the evidence that demonstrates that at least one BNYM employee was suspicious, several of the facts that Trustee maintains support a finding of knowledge do not necessarily suggest that Sentinel was misusing customer assets.
Id. at 891 (emphasis added). Again, if the Bank knew that Sentinel insiders were misusing the loan proceeds, then how could the Bank “rely on representations and warranties” made by Sentinel? True, the Bank had experienced a previously “unremarkable ten-year relationship with Sentinel.” Id. But knowledge that insiders were misusing corporate funds should have provided the Bank with more than enough reasons to distrust any representations and warranties made by Sentinel.
These inconsistencies in the district court’s opinion regarding the extent of the Bank’s knowledge before Sentinel’s collapse lead to further inconsistencies regarding the mental state of Bank employees. If Bank employees knew that Sentinel insiders were misusing loan proceeds, then it certainly suggests that Bank employees (at the very least) turned a blind eye to the rest of Sentinel’s misconduct. And yet the district court concludes that Grede “failed to prove that BNYM was deliberately indifferent to Sentinel’s alleged fraud.” Id. at 887. Even if we were to accept the district court’s conclusion that Bank employees were not deliberately indifferent to Sentinel’s misconduct (which,, of course, requires- overlooking the inconsistencies regarding the extent of the employees’ knowledge), we run into further inconsistencies regarding the employees’ mental states. The district court’s opinion appears to waffle back and forth between characterizing their mental states as negligent and as reckless. Two excerpts from the opinion illustrate our point well. Toward the end of its discussion of equitable subordination, the district court remarks:
BNYM claims it did ongoing diligence, focused on Sentinel’s creditworthiness in an effort to ensure repayment and noticed nothing. But the question of whether Sentinel had the right to pledge the collateral certainly goes to the heart of whether BNYM was adequately secured. Even if the Bank was solely concerned with protecting its own interests, a diligence process that excludes such a verification seems to be ineffective and reckless in light of the facts of which the Sentinel team at the bank was aware.
Id. at 892 (emphasis added). This statement indicates that the Bank employees were reckless in their failure to detect Sentinel’s misconduct. But only a few pages later, the district court suggests that the Bank employees’ failure to detect Sentinel’s misconduct did not rise to the level of recklessness:
The fact that they would have been better bankers if they had made a more rigorous inspection of Sentinel’s operations or its reporting is not enough to hold BNYM liable. If some degree of negligence were enough to establish inequitable conduct, the result might be different.
Id. at 894.
Because of these inconsistencies throughout the opinion, we are under[672]*672standably dubious of the district court’s ultimate conclusion that the Bank’s conduct did “not rise to the level of egregious and conscience-shocking, and its claim should not be subordinated.” Id. at 898. Thus, before the district court can reach any conclusion regarding the nature of the Bank’s conduct, we believe that the district court must clarify two critical issues:
1. What exactly did BNYM know before Sentinel’s collapse? Did BNYM know that Sentinel was engaged in misconduct of any kind (including abuse of the loan proceeds)?
2. Was BNYM’s failure to investigate Sentinel before its collapse merely negligent? Or was it reckless? Or was it deliberately indifferent?
Once the district court clarifies these two points on remand, it can then revisit the ultimate issue of whether the Bank’s claim merits equitable subordination.
C. Voiding the Contracts
The district court dismissed under Rule 12(b)(6) the claim that Sentinel’s contracts with the Bank were inherently illegal. See 11 U.S.C. § 506(d) (voiding liens securing disallowed claims). We review this dismissal de novo. Tamayo v. Blagojevich, 526 F.3d 1074, 1081 (7th Cir.2008).
The district court correctly dismissed this claim because the agreements were not the cause of Sentinel’s under-segregation. The contracts did not require either Sentinel or the Bank to do anything illegal, nor did they encourage either party to engage in illegal activity. The contracts’ provisions requiring Sentinel to release all third-party claims when funds were desegregated were not inherently unlawful because segregated funds could be deposited elsewhere Ain the normal course of business” to settle trades. 7 U.S.C. § 6d(a)(2); see also 17 C.F.R. § 1.23 (stating that § 6d(a)(2) does not prohibit an entity from withdrawing segregated funds “to the extent of its actual interest”); 17 C.F.R. § 1.29 (an FCM may receive and retain “as its own any increment or interest resulting” from investments). Furthermore, Grede fails to point to any evidence suggesting that the contract between Sentinel and the Bank was connected with an illegal scheme or plan.
Just because the parties to a contract have engaged in illegal behavior does not mean the contract itself is intrinsically illegal. Nor does “the defense of illegality ... come into play just because a party to a lawful contract ... commits unlawful acts to carry out his part of the bargain.” N. Ind. Pub. Serv. Co. v. Carbon Cnty. Coal Co., 799 F.2d 265, 273 (7th Cir.1986) (citations omitted). Consequently, we agree with the district court that the defense of illegality is inapplicable to the contracts between Sentinel and the Bank of New York.
III. Conclusion
For the foregoing reasons, we Affirm the judgment of the district court with respect to Grede’s illegal contract claim. However, we Reverse the decision of the district court with respect to Grede’s fraudulent transfer and equitable subordination claims, and we Remand the case back to the district court for further proceedings on these two claims that are consistent with this opinion. Circuit Rule 36 shall not apply on remand.