*475GIBBONS, J., delivered the opinion of the court, in which ROGERS, J., joined. MERRITT, J. (pp. 497-500), delivered a separate opinion concurring in part and dissenting in part.
OPINION
JULIA SMITH GIBBONS, Circuit Judge.Defendant-appellant David Geisen appeals his conviction on three counts of concealing a material fact and making a false statement to the Nuclear Regulatory-Commission (“NRC”) in violation of 18 U.S.C. §§ 1001 and 2. On appeal, Geisen argues that there was insufficient evidence to support his convictions and that the district court erred by giving a deliberate ignorance instruction and denying a motion to admit evidence of Geisen’s rejection of a pre-indictment deferred prosecution agreement. For the following reasons, we find that there was sufficient evidence to support each of Geisen’s convictions and that the district court did not err in its instruction or exclusion of evidence. Therefore, we affirm.
I. Factual and Procedural Background
This case arises out of an incident that occurred in 2001 at the Davis-Besse Nuclear Power Station (“Davis-Besse” or “the plant”), which is located on the shores of Lake Erie near Toledo, Ohio, and is owned and operated by FirstEnergy Nuclear Operating Company (“FENOC”). Geisen began work at the plant in 1988 and, by 2000, was manager of design basis engineering. After a safety incident at a similar plant prompted the NRC to require inspections at all like plants by the end of 2001, FENOC successfully petitioned the NRC to permit Davis-Besse to operate without interruption and thus delay inspection until a scheduled refueling shutdown in spring of 2002. Geisen’s role in preparing the documents that DavisBesse submitted to the NRC and presentations given to NRC officials in furtherance of the delayed inspection gave rise to his indictment on and subsequent conviction of three counts of concealing a material fact and making a false statement to a United States agency. During the delayed inspection, Davis-Besse found five cracked nozzle heads and a football-sized cavity caused by boric acid erosion in the head of the reactor. The finding prompted NRC investigations into previous plant inspections and, eventually, the prosecution of Geisen, systems engineer Andrew Siemaszko, and independent contractor Rodney Cook. A second engineer, Prasoon Goyal, and three other Davis-Besse employees signed deferred prosecution agreements.
A. Davis-Besse Nuclear Power Station
Davis-Besse is a two-loop, pressurized water reactor that is composed of a large cylindrical chamber filled with coolant water (“the Reactor Pressure Vessel” or “RPV”). Uranium rods at the core of the vessel fuel the nuclear reaction that heats the coolant water. The nuclear reaction is controlled by introducing boric acid and/or control rods into the reactor vessel. The control rods are inserted through sixty-nine penetration nozzles (tubes that are approximately four inches in diameter) that penetrate through the head of the reactor (approximately ten feet in diameter) into the reactor chamber. There is a gap between the RPV head and reflective metal insulation that encloses closure flanges and studs. The gap is narrowest at the top of the head, where it is only two inches wide. Control rod drive mechanisms (“CRDMs”) allow the operators to lower the rods into the reactor to control the rate of the nuclear reaction, and, thus, the energy output. The nozzles are welded onto the vessel head using a J-groove *476on the underside of the steel head, which is 6.5 inches thick.
The internal walls of the RPV and the underside of the RPV head are covered in non-corrodible stainless steel, but the RPV and the external components are made of carbon steel, which is corrodible by the boric acid in the coolant water if it escapes the RPV. This can happen when the coolant water leaks through the flanges that connect the CRDMs to the nozzles above the RPV head. Davis-Besse had a history of flange leakage and developed the Boric Acid Cdfrosion Control Procedure (“BACCP”), which it implements during inspections, to address this problem.
Davis-Besse operates in two-year fuel cycles and, therefore, shuts down the reactor only during the biennial refueling outages (“RFOs”). Davis-Besse was scheduled to conduct RF013 (the thirteenth RFO conducted at Davis-Besse) in April 2002. In addition to permitting refueling, the RFOs are the primary opportunity for inspections and maintenance that cannot occur while the reactor is in operation. The RFOs at issue in this case are RFO 10 (1996), RFO 11 (1998), and RFO 12 (2000). During an RFO, in order to visually inspect the nozzles and the RPV head, operators must insert a camera through a series of eighteen “weep holes” that are five by seven inches in size and that line the bottom of the RPV head above the head flange connecting the RPV head to the RPV. Because of the limited accessibility of the camera, it is impossible to visually inspect the very top of the RPV head and the nozzles located there. Siemaszko was in charge of inspecting and cleaning the RPV head during RF012 in 2000. Goyal oversaw this task during RFO10 in 1996 and reviewed the inspection reports following RFOll and RF012. Another engineer, Peter Mainhardt, supervised inspection and cleaning during RFOll in 1998. As of 2001, Goyal and Siemaszko continued to work at Davis-Besse as engineers, and Mainhardt worked for FENOC as an independent contractor preparing for RFO 13.
The 1996 RPV head inspection lasted only one hour due to limitations on the technicians’ exposure to radiation. During that inspection, Goyal directed two technicians who were moving a camera on a pole across the vessel head. He watched on a monitor and narrated the camera location based on the “stud hole” numbers (the numbers on the studs between the weep holes). The nozzles were not numbered, so this is the only way to determine and document the condition of each nozzle based on the camera visual. Goyal, in testimony and in a Potential Condition Adverse to Quality report (“PCAQ”) submitted to superiors after RFO 10, estimated that he was able to inspect fifty or sixty percent of the head area in 1996 and noted that it was difficult to estimate the amount of boron deposit on the head because of the limited visual inspection. In the PCAQ, Goyal attributed the boron deposits to flange leaks. The PCAQ also noted several deposits ranging in color from white to brown to rust. In both the PCAQ and in testimony, Goyal noted that the boron deposits and limited visual access prevented full implementation of the BACCP. Consequently, in the PCAQ, Goyal suggested modifications to the RPV head that would permit better access, such as installing access doors. No such modifications were ever made.
At trial, the government’s expert witness, Dr. James Davis, described photographs of the 1998 inspection, noting “rust-colored boric acid deposits coming out of the ... [weep] holes” and “boric acid deposits around the closure studs.” He also stated that “[t]here were several other [indicators of leakages]. One of them was of containment air coolers were getting *477clogged, fouled with boric acid deposits.” The RFOll PCAQ, signed by Goyal, stated that “most of the head area was covered with an uneven layer of boric acid along with some large lumps of boric acid.” The deposits were again attributed to flange leakage. That PCAQ referred back to the RFOIO PCAQ and the need for corrective action. The 1998 PCAQ also stated that “[t]he reactor vessel head was cleaned as best as we can” and noted that the cleaning was video recorded.
Siemaszko conducted RF012’s RPV head cleaning. The deposits prevented insertion of the camera into five of the weep holes and visually impaired inspection through other weep holes. The deposits also required more elaborate cleaning maneuvers than previous inspections, which had used a vacuum cleaner to remove boron deposits. In 2000, Siemaszko directed the technicians to spray hot, distilled water onto the RPV head to loosen the deposits and to use bars to knock off chunks of deposits and to flush them out through the weep holes. One of the members of the cleaning crew testified that the amount of boron deposits visible in RF012 was “unlike” any he had seen in previous RFOs and that the deposits left on the RPV head after the cleaning were of great concern to those planning RF013. Greg Gibbs, a consultant brought to Davis-Besse to prepare for RF013, reviewed the cleaning tapes of RFO 12 and testified that, although there were “large areas ... that were cleaned to bare metal[,] ... as you neared the top of the rear insulation where the two-inch gap exists ... there were areas where there were considerable boric acid deposits, in some cases even solid up to the mirror insulation.” Geisen told an NRC investigator in 2002 that he had read a report by Gibbs sometime after October 11, 2001, that discussed Gibbs’s findings in reviewing the RF012 inspection, including that the RPV head had “boric acid deposits of considerable depth.” The RF012 PCAQ again attributed the increased boron accumulation to flange leakage.
In a 2000 PCAQ, Siemaszko noted that the RPV head should be “free of boron deposits” to adequately inspect the nozzles in accordance with an NRC letter requiring plants to inspect the CRDMs adequately. Siemaszko put the RPV head on a restraint that required action before the plant was put back into operation. Geisen removed the restraint, however, stating that the RPV head would be cleaned of all boron deposits before it was put online. It was not.
B. NRC Bulletin 2001-01
In 2001, small “popcorn” deposits of boric acid were found at the nozzle penetrations of the reactor at the Oconee Nuclear Station in South Carolina, a nuclear plant of similar design to Davis-Besse. Earlier nozzle cracks had been lengthwise, but the 2001 cracks were circumferential (around the nozzle), and one was above the J-groove weld and within the “pressure boundary.” This posed a risk that the nozzle would blow out of the vessel head and cause significant loss of coolant and structural threats, including possible plant safety failure. In the early 1990s, the NRC determined that nozzles were susceptible to “stress corrosion cracking” on the nozzles and on the welding but determined that the cracks did not pose an imminent safety threat because the NRC presumed that any leakage would be readily apparent before threatening the structural integrity of the reactor or catastrophic failure. The leakages occur when coolant escapes the containment vessel within the reactor and either exits the reactor or comes into contact with the hot vessel head. The result is that the coolant flashes to steam and the boric acid within the coolant fluid is left as a deposit on the *478reactor head near the leak. In 1997, the NRC advised licensees of this type of reactor to develop programs to periodically inspect the vessel head penetrations and look for cracks but, because it was not yet aware of the problem, did not warn about the link between popcorn deposits and circumferential cracking.
FENOC was aware of the risks of circumferential cracking before 2001 because it was a member of an owner’s group that addresses problems at plants designed by Babcock and Wilcox, which designed both Oconee and Davis-Besse. Geisen was Davis-Besse’s representative to this group. At trial, Geisen testified that he was first involved with nozzle-crack issues in late 2000, after the first cracks were found at Oconee, and that he had given several presentations on the subject beginning in the spring of 2001. Goyal had also sent numerous emails to Geisen and others warning that “head cleaning during outages should be a top priority” and, after the circumferential crack at Oconee was discovered, Goyal sent an email stating that the five nozzles at Davis-Besse located at the center of the RPV head were manufactured in the same way as were all of the cracked Oconee nozzles. However, in June 2001, Geisen approved a memorandum, prepared by Goyal, that concluded that Davis-Besse could postpone inspection of the nozzles until RF013. The memorandum acknowledged that significant boron leakage from flanges had impeded “detailed inspection of CRDM nozzles” during RF012 but calculated that “it would take approximately 2.5 additional years of operation for Davis-Besse to observe the same degradation” as occurred at Oconee.
After receiving notice of the Oconee cracking, the NRC altered its assessment of the risks of even small boron deposits on reactor heads. In light of the Oconee incident and similar experiences in the French nuclear industry, on August 3, 2001, the NRC issued NRC Bulletin 2001-01 (“NRC 2001-01” or “the Bulletin”), entitled “Circumferential Cracking of Reactor Pressure Vessel Head Penetration Nozzles.” The Bulletin outlined which plants had a “high susceptibility” to nozzle stress cracking and the NRC’s criteria indicated that Davis-Besse was among them. The Bulletin also requested information from affected nuclear power stations such as Davis-Besse. The Bulletin stated that such plants “need to use a qualified visual examination of 100% of the ... nozzles,” that the inspection “should be able to reliably detect and accurately characterize leakage from cracking,” and that “the effectiveness of the ... examination should not be compromised by the presence of insulation, existing deposits on the RPV head, or other factors that could interfere with the detection of leakage.” Due to the risks, the NRC wanted all high-risk plants such as Davis-Besse to shut down and conduct a complete inspection for nozzle cracks by December 31, 2001. Because of the costs involved in an early shutdown, Davis-Besse wanted to continue operation until its scheduled RF013 in April 2002.
The Bulletin required plants to provide detailed information about susceptibility to cracking and previous inspections within thirty days. As part of that information, the NRC directed high-risk plants that, “[i]f [the plant’s] future inspection plans do not include performing inspections before December 31, 2001, [the plant must] provide [the] basis for concluding that the regulatory requirements discussed in the Applicable Regulatory Requirements section will continue to be met until the inspections are performed.” Section l.d. required all such plants to provide:
[A] description of the [vessel head penetration] nozzle and RPV head inspec*479tions (type, scope, qualification requirements, and acceptance criteria) that have been performed at your plant(s) in the past 4 years, and the findings. Include a description of any limitations (insulation or other impediments) to accessibility of the bare metal of the RPV head for visual examinations.
C. Davis-Besse’s Representations to the NRC
In accordance with federal regulations governing the nuclear industry, DavisBesse was obligated to respond to the NRC Bulletin with “written statements, signed under oath or affirmation.” 10 C.F.R. § 50.54(f); see also 42 U.S.C. § 2011 et seq. Federal regulations also require that all information provided to the NRC “be complete and accurate in all material respects.” 10 C.F.R. § 50.9(a). Davis-Besse hired Cook to coordinate the response to NRC 2001-01. Between September 4 and November 30, 2001, FENOC submitted a series of serial letters (“SLs”) containing the information requested in the Bulletin. Various conference calls and meetings between FENOC employees and the NRC also took place between September 4, 2001, and December 4, 2001, when the NRC finally permitted Davis-Besse to continue operation until an earlier shutdown for RF013 in February 2002. The five letters at issue in this case and charged to contain false statements in the indictment against Siemaszko, Geisen, and Cook are: SL 2731, September 4, 2001 (count 1); SL 2735, October 17, 2001 (count 2); SL 2741, October 30, 2001 (count 3); SL 2744, October 30, 2001 (count 4); and SL 2745, November 1, 2001 (count 5). Count 1 also included allegations of concealment of material facts in the other serial letters and during meetings between FENOC staff — including Geisen — and the NRC. Geisen was convicted on the first, third, and fourth counts of the indictment.
In approving Davis-Besse’s continued operation until RFO 13, the NRC relied on all of the serial letters:
Based on the information provided in your responses [dated September 4, 2001, as supplemented by letters dated October 17, October 30, November 1, and November 30, 2001] and the information available to the staff regarding the industry experience with VHP nozzle cracking, the staff finds that you have provided sufficient information to justify operation until February 16, 2002, at which time you will shut down the [plant] ... and perform VHP nozzle inspections as discussed in your letter dated November 30, 2001. The commitments contained in your letter dated November 30, 2001, were integral to the staffs finding.
The serial letter submitted on November 30, 2001, SL 2747, was not readily discoverable in the record.
FENOC’s first submission to the NRC in response to NRC 2001-01 was SL 2731 on September 4, 2001. Siemaszko was tasked with reviewing the inspection tapes from previous RFOs and providing information in response to NRC 2001-01’s section l.d. inquiry, Cook was in charge of putting together the information, and Goyal was to review the submission. Siemaszko wrote the first draft, which stated that the guidance procedure predating BACCP was used in RFO 11 and RF012, that “[t]he head cleaning was limited by the opening size of the weep holes,” and that, during RF012, “[n]o evidence of nozzle leakage was detected. 95% of the nozzles were inspected.” Goyal questioned the ninety-five-percent assertion given the amount of boron visible on the top of the RPV head during RF012, and Siemaszko subsequently sent another draft asserting *480that “[n]o visible evidence of nozzle leakage was detected!, m]ajority of nozzles were inspected,” and stating that the procedure used was the BACCP. Later, after Cook questioned the meaning of “majority,” Siemaszko stated that ninety percent of the nozzles had been inspected. Goyal expressed concern regarding the ninety-percent claim and the assertion in the draft that all of the CRDMs were inspected given the amount of boric acid deposits obstructing the view.
Nevertheless, the final letter included the statement that “a gap exists between the RPV head and insulation, the minimum ... is approximately 2 inches, and does not impede visual inspection.” The letter also asserted that Davis-Besse’s BACCP procedure had been utilized in both inspections and that “[t]he scope of the visual inspection was to inspect the bare metal RPV head area that was accessible through the weep holes to identify any boric acid leaks/deposits.” SL 2731 also described the boron deposits discovered during the 1998 inspection as an “uneven layer of boric acid deposits scattered over the head ... [and] some lumps of boron, with the color varying from brown to white.” Of the 2000 inspections, SL 2731 noted that “[s]ome boric acid crystals had accumulated on the RPV head insulation beneath the leaking flanges. These deposits were cleaned (vacuumed),” that “[i]nspection of the RPV head/nozzles area indicated some accumulation of boric acid deposits,” and that the RPV head area was cleaned with demineralized water to the greatest extent possible. Referencing the review of the videotaped 1998 and 2000 inspections conducted in May 2001, following Oconee, SL 2731 also noted that “indications such as those that would result from RPV head penetration leakage [like at Oconee] were not evident.” SL 2731 also asserted that a full inspection, unimpeded by boric deposits, would take place during RFO 13.
Each serial letter sent to the NRC included a “green sheet,” which is a cover document listing FENOC employees who contributed to and/or reviewed the document before it was sent to the NRC. There is space for each listed employee to sign and date the letter. Geisen, as design engineering manager, signed and dated the green sheet both on his own behalf and on behalf of his supervisor, Steve Moffltt, who was the director of technical services. Goyal testified that he was uncomfortable signing the green sheet because it misrepresented how thorough the prior inspections and cleanings were, but he eventually did so.
On September 28, 2001, the NRC contacted Davis-Besse to urge it to reconsider its approach to its NRC 2001-01 submissions and to suggest shutting down the plant before December 31, 2001, in order to conduct a proper inspection of the nozzle heads. At trial, Moffltt testified that “December versus April became this issue of great discussion” at Davis-Besse because the difference in consequences of an outage in 2001-several months before the end of a fuel cycle — and at the completion of the cycle would be “quite severe.” He stated:
It wasn’t just this outage; it was for the next 20, 40 years you would not be operating at your full tank of gas as you saw it. Then there was fuel [(its availability was questionable)]; there was cost; there was certainly morale, a Christmas outage; there was dose [of radiation] ... [;] our own sense of confidence.
During an interview with an NRC investigator following RFO 13, Geisen stated that the site vice-president was “very upset” by the September 28 call from the NRC, *481which prompted “all sorts of new work activity.”
At this point, management-level personnel, and Geisen in particular, began to take a more active role in the NRC negotiations. Geisen took part in a conference call on October 3, 2001, during which he represented — incorrectly—that 100 percent of the RPV head had been inspected during the RFOs and that boric acid deposits only impeded visual inspection of five or six nozzles. On October 11, 2001, Geisen and other managers gave a slide presentation to NRG staff. Geisen compiled the information for the “facts” slides, but it is unclear whether Geisen or Moffitt presented them. One crucial slide stated that “[a]ll CRDM penetrations were verified to be free from ‘popcorn’ type boron deposits using video recordings from 11RFO or 12RFO.” Moffitt testified that Geisen later determined that the 100-percent statement was only attributable to RFO10 rather than to the later RFOs and decided to correct the error in subsequent submissions.
On October 17, 2001, FENOC sent SL 2735 to the NRC to supplement SL 2731. Geisen, as a “responsible manager,” initialed and dated the green sheet for this submission. SL 2735 contained a table detailing the status of each nozzle at each inspection (“nozzle inspection table”). The table indicated whether each nozzle had been recorded and whether leaks were apparent on each nozzle. After the NRC’s request for more information following SL 2731, Geisen had asked Siemaszko to review the inspection videos and to prepare the table.
After receiving Siemaszko’s draft table charting the 1998 and 2000 inspections, Geisen told Siemaszko to include the 1996 inspection. Because he had never seen the RPV head in 1996, Siemaszko relied on information from others to complete the table. For the 1998 and 2000 inspections, each nozzle had one of the following notations: (1) “no leak observed,” indicating that a visual inspection was sufficient and no video record was needed; (2) “no leak recorded,” indicating that the nozzle inspection was recorded on the video; or “flange leak evident,” indicating that the nozzles were not visible due to boric acid deposits.
During the NRC investigation into Davis-Besse in 2002, Geisen told investigators that he was responsible for supervising Siemaszko’s work on the nozzle inspection table and, according to testimony by the investigator, “[Geisen] said that during ... an early October time frame, ... he had viewed portions [of the videos] of the 1996, 1998 and the 2000 reactor vessel head inspections.” The version of SL 2735 submitted to the NRC contained the nozzle inspection table as Attachment 2, with a footnote to the 1996 inspections stating that “the entire RPV head was inspected. Since the video was void of head-orientation narration, each specific nozzle view could not be correlated.” The letter also stated that “50 of -69 nozzles” were “viewed” in 1998, “45 of 69” were “viewed” in 2000, and that some nozzles were not viewed in 2000 because they were “obscured by boric acid crystal deposits that were clearly attributable to leaking ... flanges from the center CRDMs.” The letter noted that the visual inspections in 1996, at which time sixty-five of the sixty-nine nozzles were inspected, and in 1998 and 2000 “consisted of a whole head visual inspection” as required by the BACCP. The document also asserted that none of the videos indicated “boric acid chrystal deposits that would have been attributed to leakage from the CRDM nozzle penetrations.”
Based on the assertion that all nozzles were leak-free prior to RFO10, as demon*482strated in the table, FENOC conducted a risk analysis that determined that the earliest a crack could have developed was May 1996, after RFOIO concluded. In the worst-case scenario, that crack would take seven-and-one-half years to grow to beyond a safe size, and, therefore, DavisBesse could safely operate until RFO 13. This risk analysis formed the basis of Davis-Besse’s representations to the NRC that a delayed inspection was safe.
Despite the detailed nozzle inspection table, the NRC was still not satisfied that Davis-Besse could operate safely until the scheduled outage in April 2002. Consequently, on October 24, 2001, Geisen again presented slides to the NRC, including one that stated that “the inspection results afford us assurance that all but 4 nozzle penetrations were inspected in 1996” and that “no penetration leakage was identified.” The NRC, however, remained unconvinced.
On October 30, 2001, FENOC submitted two further serial letters to the NRC, both of which contained the nozzle inspection table. SL 2741 included a risk analysis and reiterated that, taken together, the inspections in 1996, 1998, and 2000 constituted a “whole head visual inspection” of the “bare head” in accordance with the BACCP. SL 2744 contained still photographs taken from the inspection videos. Siemaszko provided the “representative” photographs, and Geisen wrote the captions. Geisen initialed and dated the green sheets for both letters.
Geisen testified that he asked Siemaszko to collect “representative” photographs and drafted the captions based on previous conversations — unrelated to the drafting of SL 2744 — that he had with Siemaszko about the inspections. Geisen also testified that he did not watch the videos in their entirety before compiling the photographs. The government entered evidence, however, suggesting that Geisen had viewed the videos in August 2001 and, at least partially, in preparation for submitting SL 2735.
In an introduction to the 1996 photographs, a caption characterizes the photographs as “representative” and the head as “relatively clean and afford[ing] a generally good inspection.” The caption to a photograph showing boric acid deposits at the top of the RPV head states that the deposits could not be removed by mechanical cleaning because of their “location,” but were “in the vicinity of previous leaking flanges” and “verified not to be active or wet.” Geisen told investigators that Edward Chimahusky, then an engineer in charge of coolant systems at Davis-Besse, provided the information for that caption, but Chimahusky testified that he was not involved in the response to NRC 2001-01 and that Geisen never consulted him regarding the captions or response to NRC 2001-01. Chimahusky also testified that he had only inspected the flanges on the outside of the reactor head and had not inspected the interior. The photographs included in the letter as “representative” did not show any of the more significant piles of boric acid deposits that the videos contained.
On November 1, 2001, FENOC submitted SL 2745, which contained a “plant specific assessment” expanding on the risk assessment provided in SL 2741. However, despite these submissions, the NRC continued to deny permission to delay the full-head inspection required by NRC 2001-01. In an effort to convince the NRC that delaying inspection was safe, Geisen presented excerpts of the prior inspection videos to NRC staff on November 8, 2001. According to testimony at trial by Dr. Allen Hiser, one of the NRC staff attending the presentation, Geisen showed excerpts of the 1996 video “to confirm that *483the head ... was in good condition in 1996” but did not show certain segments of the video that showed large deposits on the RPV head. Hiser testified:
In retrospect, the good portions I think is what we reviewed. Mr. Geisen had control of the remote ... and ... he would fast-forward and jump to various places in the tapes, and we would review maybe for a minute or five minutes just looking at the general condition of the head that was visible, and then we’d go maybe forward.
Geisen testified, however, that he had shown the entire 1996 video. Hiser also testified that when he reviewed the same tapes during the NRC investigation in 2002, he saw “a lot more boron than we had expected ... which was inconsistent really with anything that we had been provided previously.” Hiser stated that he “ha[d] no idea” whether Geisen intended to skip over the parts of the video showing significant boron buildup but did know that the excerpts that he showed were not representative. Hiser also stated that Geisen showed portions of the 1996 and 1998 videos but no portion of the 2000 inspection videos. According to Hiser, Geisen said “if you think this tape is bad, the 2000 tape is even worse, so I won’t bother to show it to you.” Both Geisen and Hiser agreed that Geisen was unable to narrate the videos, and Geisen testified that he had not previously viewed the videos in their entirety and had no time to prepare. Geisen testified that because he was frustrated with his inability to narrate, he arranged for Siemaszko to meet with the NRC to review the videos. Siemaszko did meet with NRC staff on November 14, 2001, to provide assurances that the previous inspections had been sufficient.
As the December 31, 2001, deadline for inspection approached, FENOC managers — including Geisen — met with NRC staff on November 28, 2001, to discuss whether Davis-Besse would have to close. At that meeting, FENOC made additional commitments to expand the scope and bring forward the timing of RF013, including proposing an earlier shutdown date of February 16, 2002, and promising to conduct a “100% qualified visual” and “100% [non-destructive examination]” inspection. FENOC also committed to replace the vessel head “at first available opportunity.”
D. Procedural Background
During the resulting 2002 inspection, the plant discovered a large cavity in the head of the reactor created by boric acid eroding the steel. The erosion had penetrated through the carbon steel wall, leaving only the 0.24" to 0.38" stainless steel lining of the reactor head, and was located near five cracked nozzles, four of which were at the very top of the reactor head (nozzles 1,2, 3, and 5). The cavity was discovered only by chance when one of the cracked nozzles moved. As a result of the ensuing internal investigation, Davis-Besse fired Siemaszko and Goyal in September 2002 because of their roles in providing inaccurate and misleading information to the NRC in the serial letters.
In January 2006, a grand jury indicted Geisen, Siemaszko, and Cook on five counts of violating 18 U.S.C. §§ 1001 and 2.1 The indictment charged that, based on the statements made in the serial letters submitted to the NRC and at two public meetings, the NRC permitted Davis-Besse to operate beyond December 31, 2001. Count 1 charged that the three “did knowingly and willfully conceal and cover up, and cause to be concealed and covered up, by tricks, schemes and devices, material *484facts in a matter within the jurisdiction of the executive branch of the government of the United States, to wit, the condition of Davis-Besse’s reactor vessel head, and the nature and findings of previous inspections of the reactor vessel head.” The detailed indictment regarding count 1 listed SL 2731, the other serial letters, and various meetings with NRC authorities between September and December 2001 in which the three defendants participated in various ways. Counts 2 through 4 alleged that Geisen “did knowingly and willfully make, use, and cause others to make and use a false writing,” including: (count 2) SL 2735, containing five allegedly false statements; (count 3) SL 2741, containing five allegedly false statements; and (count 4) SL 2744, containing six allegedly false statements. Count 5 alleged that Geisen “did knowingly and willfully cause others to make and use a false writing.”
Before the indictment was returned, the government offered Geisen a deferred prosecution agreement (“DPA”) that promised that the government would “refrain from seeking an indictment or otherwise initiating criminal prosecution of ... Geisen” with respect to stated stipulated facts. The DPA required Geisen to “ad-mite ] that between September 3, 2001, and November 28, 2001, he knowingly and deliberately caused false representations to be made to the NRC in the course of attempting to persuade the NRC that [Davis-Besse] was safe to operate beyond December 31, 2001.” The DPA also stated that Geisen would waive the statute of limitations for future prosecution based on a breach of the DPA, cooperate in criminal and administrative proceedings related to the incident, and agree that the stipulated facts could be used against him in any proceeding should he breach the DPA. Geisen rejected the DPA, and he was indicted.
Geisen and Cook moved to sever their trial from that of Siemaszko. The district court granted the motion, and Geisen and Cook were tried jointly in October 2007.2 The key issue at Geisen’s trial was whether he possessed the intent required by § 1001. Geisen, who testified in his own defense, contended that although the statements were false, he did not know that they were false at the time and did not intend to deceive the NRC. As evidence of a lack of intent, Geisen filed a motion in limine seeking to introduce evidence of his pre-indictment rejection of the offered DPA. The district court denied the motion.
Among the testimony relied on heavily by the government was that of John Martin, a former NRC investigator who interviewed Geisen in 2002. Martin, relying on his handwritten notes of the interview, testified that Geisen stated that he viewed the inspection videos in August 2001 in connection with preparing for Davis-Besse’s interactions with the NRC. This contradicted Geisen’s own testimony that he had not reviewed the video tapes at the time that he wrote the captions for the photographs submitted to the NRC in SL 2744. The government also submitted into evidence numerous emails that were addressed or copied to Geisen from Goyal discussing past inspections at Davis-Besse, similarities between Oconee and Davis-Besse, and the need to modify the RPV head to permit better access during inspections and cleaning. In one such email, Goyal noted that Davis-Besse’s was the only Babcoek- and-Wilcox-designed reactor that did not have access doors on the reactor head.
To demonstrate the falsity of the statements included in the serial letters, the *485government introduced the inspection videos and summaries of the prior cleanings into evidence through the expert testimony of Melvin Holmberg. Holmberg, who conducted an audit of the inspections and created a “map” of the RPV head identifying each nozzle by number, walked the jury through the various videos. He identified which nozzles were visible during each inspection and to what extent the view of each nozzle was sufficient to enable the “qualified visual examination” (“QYE”) required by NRC 2001-01. In the diagrams he produced, he also identified which of those nozzles were designated by FENOC as “no leak observed,” i.e., “visual inspection satisfactory, no video record required,” and which were designated as affected by flange leakages. Summarizing Holmberg’s results, the government included in its brief before this court the following table illustrating how many of the nozzles were visible for inspection:
[[Image here]]
The government claimed that this was inconsistent with the assertions made in SL 2735 that “50 of 69 nozzles” were visibly inspected in 1998 and “45 of 69” were visibly inspected in 2000, although the serial letter did not differentiate between QVE and “viewed.”
At the end of the trial, the government asked for and was granted a jury instruction on deliberate ignorance. After three days of deliberations, the jury reached a verdict of acquittal as to Cook and informed the district court that it had reached only a partial verdict as to Geisen. After hearing an Allen charge, the jury returned a guilty verdict on counts 1, 3, and 4. The district court denied a motion for judgment of acquittal and new trial, noting that, “[a]lthough a close case, the evidence presented, including testimony from the Defendant himself, when viewed cumulatively, constitutes sufficient direct and circumstantial evidence upon which a reasonable jury, utilizing the standard ‘beyond a reasonable doubt/ could have based a finding of knowledge and intent.” Geisen was sentenced to three years of probation for each count, to run concurrently, and was fined $7,500, directed to perform 200 hours of community service, and prohibited from working in the nuclear industry during his period of probation. Geisen timely appealed.
II. Deliberate Ignorance Instruction
We review challenges to a district court’s jury instruction for abuse of discretion. United States v. Prince, 214 F.3d 740, 761 (6th Cir.2000). “A trial court has broad discretion in crafting jury instructions and does not abuse its discretion unless the jury charge ‘fails accurately to reflect the law.’ ” United States v. Ross, 502 F.3d 521, 527 (6th Cir.2007) (citations omitted). Thus, we may reverse the jury’s conviction “only if the instructions, viewed as a whole, were confusing, misleading, or prejudicial.” United States v. Harrod, 168 F.3d 887, 892 (6th Cir.1999) (internal citations and quotation marks omitted).
We have stated that a deliberate ignorance instruction is warranted to “prevente ] a criminal defendant from escaping conviction merely by deliberately closing *486his eyes to the obvious risk that he is engaging in unlawful conduct.” United States v. Gullett, 713 F.2d 1203, 1212 (6th Cir.1983). However, we have cautioned that this instruction should be used sparingly because of the heightened risk of a conviction based on mere negligence, carelessness, or ignorance. See United States v. Mari, 47 F.3d 782, 787 (6th Cir.1995) (warning courts not to use the instruction “indiscriminately”); see also Pattern Criminal Jury Instructions for the Sixth Circuit § 2.09. A deliberate ignorance instruction is properly given, therefore, when there is evidence supporting an inference of deliberate ignorance. See United States v. Lee, 991 F.2d 343, 351 (6th Cir.1993).
The district court properly instructed the jury that it could only find Geisen guilty under a deliberate ignorance theory if it was “convinced beyond a reasonable doubt that the defendant was aware of a high probability that the submissions and presentations to the NRC concealed material facts ... or included false statements.” See id. at 350-51 (upholding the use of the same instruction). The district court further cautioned the jury that “[cjarelessness, or negligence, or foolishness on [the defendant’s] part is not the same as knowledge and is not enough to convict.”
Geisen also argues that the instruction confused and misled the jury and permitted it to convict on the basis of negligence rather than criminal intent to deceive. Geisen argues that the risk of confusion is greater than usual in this case because the government entered considerable evidence that Geisen was negligent in preparing and reviewing the submissions to the NRC. We find Geisen’s argument unavailing. In giving the instruction to the jury, the district court was very careful to use Pattern Jury Instruction 2.09. The court also gave a limiting instruction. We have held that Pattern Jury Instruction 2.09 is an accurate statement of the law. Id. at 351. And, in Mari, we found that cautionary language such as that used by the district court in this case “forecloses the possibility of th[e] error” that a conviction is improperly based on negligence or carelessness. 47 F.3d at 785.
Quoting the Tenth Circuit, Geisen also argues that it is illogical to give the deliberate ignorance instruction since the government maintains that the evidence supports actual knowledge based on Geisen’s knowledge of the contents of the Goyal emails. See United States v. Francisco-Lopez, 939 F.2d 1405, 1410 (10th Cir.1991) (“If evidence proves the defendant actually knew an operant fact, the same evidence could not also prove he was ignorant of that fact.”). Mari forecloses this argument because it held that improperly giving the “deliberate ignorance” instruction is at most harmless error when the prosecution presented sufficient evidence of actual knowledge. 47 F.3d at 786 (citing Griffin v. United States, 502 U.S. 46, 55-56, 112 S.Ct. 466, 116 L.Ed.2d 371 (1991) (holding that giving an instruction based on unsupported grounds is harmless as a matter of law)); see also United States v. Springer, 262 Fed.Appx. 703, 706 (6th Cir.2008) (finding the same argument “hard to swallow” given the defendant’s argument at trial that he had no actual knowledge). In so holding, the Mari court
recognized that [it] must assume that the jury obeyed the language of the district court’s instructions. The words of the instruction required the jury to find beyond a reasonable doubt that the defendant was deliberately ignorant before it could convict on that ground. Therefore, even if there had been insufficient evidence to support a deliberate ignorance instruction, we must assume that the jury followed the jury charge *487and did not convict on the grounds of deliberate ignorance. Thus, another theory must have formed the basis for the conviction.
United States v. Monus, 128 F.3d 376, 390-91 (6th Cir.1997) (citing Mari, 47 F.3d at 785-87) (internal citations omitted).
Because the jury instruction given by the district court in this case was not an incorrect statement of the law but rather at worst — if we take Geisen’s arguments at face value — “one that is simply not supported by the evidence,” it was not prejudicial to Geisen. See Mari, 47 F.3d at 786 (quoting Griffin, 502 U.S. at 59, 112 S.Ct. 466).3 The district court, therefore, did not abuse its discretion by giving a “deliberate ignorance” instruction and that no prejudice to Geisen resulted from that instruction. Moreover, we find below in discussing the sufficiency of the evidence that Geisen’s convictions can be upheld under an actual knowledge theory. Therefore, any possible error in giving the deliberate ignorance instruction was harmless. Id. at 786.
Even if Geisen’s convictions cannot be sustained under an actual knowledge theory, the evidence was sufficient to demonstrate that he acted with deliberate ignorance. In order to constitute a violation of § 1001, a false statement must be made to or a material fact concealed from the NRC “knowingly and willfully.” 18 U.S.C. § 1001. Consequently, “[t]o establish a violation of § 1001, the Government must prove beyond a reasonable doubt that the statement was made with knowledge of its falsity,” United States v. Yermian, 468 U.S. 63, 64, 104 S.Ct. 2936, 82 L.Ed.2d 53 (1984), and an “intent to deceive,” United States v. Ahmed, 472 F.3d 427, 433 (6th Cir.2006). Geisen contends that with respect to all three convictions, the government failed to prove either knowledge or intent to deceive. He argues that in his supervisory role, he relied in good faith on information given to him by those with first-hand knowledge in compiling and reviewing the submissions to the NRC and making presentations to NRC staff. He also argues that the government improperly imputes to him knowledge of anything that FENOC knew as well as FENOC’s motive.
Geisen argues that the government’s approach to deliberate ignorance constitutes conviction on the basis of negligence— what Geisen should have■ known — rather than because he “consciously attempted to escape confirmation of conditions or events he strongly suspected to exist.” United States v. Skilling, 554 F.3d 529, 548 (5th Cir.2009), cert. granted, — U.S.-, 130 S.Ct. 393, 175 L.Ed.2d 267 (2009) (quoting United States v. Lara-Velasquez, 919 F.2d 946, 951 (5th Cir.1990), and declining to find the instruction improperly given). We disagree.
The government identified several representations that it argues “demonstrate a deliberate disregard for the truth,” thereby suggesting that Geisen was “deliberately avoiding culpable knowledge.” First, Geisen testified that he compiled information for slides presented to the NRC at a meeting on October 11, 2001. One of the slides represented that “[a]ll CRDM penetrations were verified to be free from *488‘popcorn’ type boron deposits using video recordings from 11 RFO or 12RFO.” Geisen testified that the information for that slide came from Siemaszko’s review of the tapes, but acknowledged that Siemaszko had not completed his nozzle inspection table by the time the slides were composed. The government argues that “this means that Geisen personally vouched for the comprehensiveness of the inspections, without any basis for doing so.” Second, Geisen testified that he did not recall ever speaking “face-to-face” with Siemaszko regarding Siemaszko’s assertion that the 1996 inspection had visualized sixty-five of sixty-nine nozzles, despite Geisen’s knowledge of the considerable deposits remaining on the RPV head and the impediments to previous inspections. Third, Geisen testified that he did not consult Siemaszko or any systems engineers while drafting the captions for the photographs submitted in SL 2744 but rather relied on his memory of “previous conversations.” Taken together with his testimony that he never reviewed the inspection videos on his own, the government argues that “[t]his left the jury to understand that Geisen interpreted images and made ciitical representations about past inspections — e.g., that certain boric acid deposits were ‘verified not to be active or wet’ — without any confirmed basis for doing so.”
In conclusion, the government presented ample evidence from which a rational jury could infer that Geisen deliberately chose not to inform himself in preparing the submissions to the NRC. Testimony and documents entered into evidence suggested that Davis-Besse’s representations to the NRC in the serial letters, meetings, and conference calls played a leading role in convincing the NRC that it would be safe to keep Davis-Besse in operation beyond December 31, 2001; that Geisen admittedly put little effort into informing himself and confirming the assertions he made to the NRC in the serial letters and in person at meetings; that there were significant and readily apparent inconsistencies between the information Geisen received from Siemaszko and others and the actual state of the RPV head; that Geisen possessed knowledge of the nature of the plant’s prior inspections from his involvement in reviewing procedures in the wake of the Oconee incident; and that the plant’s management desired to keep the plant in operation until RF013. Consequently, the district court’s instruction was not improper.
III. Sufficiency of the Evidence Claims
We review a district court’s refusal to grant a motion for judgment of acquittal and a defendant’s claim of insufficiency of the evidence de novo. See United States v. Gunter, 551 F.3d 472, 482 (6th Cir.2009) (sufficiency of the evidence claims); United States v. Kone, 307 F.3d 430, 433 (6th Cir.2002) (motions for acquittal). “[T]he relevant question is whether, after viewing the evidence in the light most favorable to the prosecution, any rational trier of fact could have found the essential elements of the crime beyond a reasonable doubt.” Jackson v. Virginia, 443 U.S. 307, 319, 99 S.Ct. 2781, 61 L.Ed.2d 560 (1979) (emphasis in original); see also United States v. Dedman, 527 F.3d 577, 592 (6th Cir.2008).
All conflicts in the testimony are resolved in favor of the government, and every reasonable inference is drawn in its favor. United States v. Bashaw, 982 F.2d 168, 171 (6th Cir.1992). In considering the claim, “we do not weigh the evidence presented, consider the credibility of witnesses, or substitute our judgment for that of the jury.” United States v. M/G Transp. Servs., Inc., 173 F.3d 584, 588-89 (6th Cir.1999) (citing United States v. Hill*489iard, 11 F.3d 618, 620 (6th Cir.1993)). This standard applies even if the evidence is purely circumstantial. See Kone, 307 F.3d at 434. Consequently, in raising a sufficiency of the evidence claim, a defendant “bears a very heavy burden.” United States v. Spearman, 186 F.3d 743, 746 (6th Cir.1999).
In order to convict a defendant for making false statements to a federal agency in violation of 18 U.S.C. § 1001, the government must prove: “(1) the defendant made a statement; (2) the statement is false or fraudulent; (3) the statement is material; (4) the defendant made the statement knowingly and willfully; and (5) the statement pertained to an activity within the jurisdiction of a federal agency.” Dedman, 527 F.3d at 598 (quoting United States v. Lutz, 154 F.3d 581, 587 (6th Cir.1998)). Only the last element is undisputed. When, as in the instant case, the indictment alleges multiple fraudulent statements for each count, this court must “uphold a conviction where there was sufficient evidence for at least one of the alleged false statements” for each count. Id. (emphasis added). After reviewing the extensive record in this case, we find that the government presented sufficient evidence to sustain Geisen’s convictions on all three counts.
A. Count 3 — Making False Statements in SL 2741
Count 3 of the indictment charged Geisen with “knowingly and willfully mak[ing], us[ing], and eaus[ing] others to make and use a false writing, that is, [SL 2741], knowing that it contained the following material statements, which were fraudulent” to the NRC in violation of §§ 1001 and 2. The allegedly false material statements were:
1. “[djuring 10RFO, 65 of 69 nozzles were viewed” ...;
2. “[i]n 1996 during 10 RFO, the entire RPV head was inspected" ...;
3. “[s]ince the [RFO10] video was void of head orientation narration, each specific nozzle view could not be correlated”
4. “[t]he inspections performed during the 10th, 11th, and 12th [RFOs] ... consisted of a whole head visual inspection of the RPV head in accordance with the [BACCP]” ...; and
5. “Hollowing 12RFO, the RPV head was cleaned with demineralized water to the extent possible to provide a clean head for evaluating future inspection results” ....
We must uphold the conviction on this count if there was sufficient evidence for a jury to convict based on any one of these five allegations. See Dedman, 527 F.3d at 598.
SL 2741, submitted on October 30, 2001, in conjunction with SL 2744, provided a risk analysis based on the assumption that a “whole head visual inspection” of the “bare head,” excepting only four nozzles, had been conducted in accordance with the BACCP in 1996. Geisen testified that he asked Siemaszko to expand the nozzle inspection table to include the 1996 inspection because he realized that the 1998 and 2000 inspections even taken together did not amount to a full visual inspection of the reactor head. Thus, Geisen was aware of the paramount importance of representing to the NRC that the 1996 inspection was complete for sixty-five of the sixty-nine nozzles. Had the inspection been less complete, the risk analysis would be inaccurate, and Davis-Besse could not assure the NRC that it had visually inspected the “entire head” as recently as 1996.
Testimony also suggested that DavisBesse’s managers, including Geisen, were under considerable pressure from their su*490periors to keep the plant in operation. Geisen’s supervisor, Moffitt, testified that FENOC was very concerned that DavisBesse continue operating as scheduled until spring 2002. After SL 2731, FENOC’s first response to NRC 2001-01, proved unsuccessful in securing the NRC’s permission to continue operations, Geisen took a more active role in coordinating and overseeing Davis-Besse’s response to NRC 2001-01. Moffitt, Geisen’s direct supervisor, testified that “December versus April became this issue of great discussion” and that the economic, technical, and morale repercussions of halting operation before the completion of the cycle would be “quite severe.” Geisen himself testified that the site vice-president was “very upset” when SL 2731 was unsuccessful, which created “all sorts of new work activity.” Evidence presented at trial that Geisen played a direct role in drafting and reviewing SL 2735 by directing Siemaszko and Goyal demonstrates that Geisen was directly involved in Davis-Besse’s efforts to convince the NRC to allow the plant to continue operating until RF013. Geisen also drafted the text of SL 2744. A rational jury, therefore, could have concluded that Geisen had a motive and intent to deceive the NRC in order to keep the plant in operation through spring 2002.
The government also presented sufficient evidence at trial for a rational jury to find that Geisen knew that SL 2741 misrepresented the success of the prior inspections and the extent of the cleaning of the RPV head. First, there was sufficient evidence for a rational jury to find beyond a reasonable doubt that Geisen knew that the inspection in 1996 had not been of the “entire head” and had not covered sixty-five of the sixty-nine nozzles as alleged in SL 2741. Furthermore, a rational jury could find that there was sufficient evidence that Geisen knew that the BACCP had not been utilized in the inspections because of the extent of the deposits noted in the 2000 PCAQ that he had reviewed and because of Goyal’s emails to him.
Second, Geisen testified that as plant representative to the Babcock and Wilcox owner’s group, he was aware of the risks of nozzle cracking and had given presentations on the risks associated with the cracking in early 2001. He was also involved in reviewing a June 2001 memorandum that stated that while the plant could operate safely until RF013, significant boron deposits that remained on the RPV head following RF012 must be addressed. Furthermore, because of his involvement in reviewing the 2000 PCAQ and canceling the operational restraint imposed by Siemaszko, Geisen was aware that the whole head had not been cleaned during the RF012 process in 2000 as represented in SL 2741. Thus, a rational jury could have determined that Geisen knew that SL 2741’s assertion that, “[f]ollowing 12RFO, the RPV head was cleaned with demineralized water to the extent possible to provide a clean head for evaluating future inspection results” was materially misleading. Finally, the jury could have credited Martin’s testimony that Geisen reviewed the inspection videos personally in August 2001, and, therefore, knew that there was narration on the 1996 video, contrary to representations that there was none in SL 2741, and that fewer than sixty-five nozzles had been visible.
Therefore, although a rational jury would have had to rely on largely circumstantial evidence to infer Geisen’s knowledge of the falsity of the statements in SL 2741, given the evidence presented, one could have found that Geisen knew that certain assertions in SL 2741 were false and incomplete and, knowing this, signed and submitted SL 2741 to the NRC with the intent to represent the past inspections as more complete than they had been. *491Therefore, there was sufficient evidence for a jury to find Geisen guilty of violating §§ 1001 and 2 as alleged in count 3.
B. Count k — Making False Statements in SL
The government’s case on count 4 is perhaps the strongest against Geisen, and we find that a rational jury could find all elements of § 1001 beyond a reasonable doubt. Count 4 of the indictment charged Geisen with “knowingly and willfully mak[ing], us[ing], and causing] others to make and use a false writing, that is, [SL 2744], knowing that it contained the following material statements, which were fraudulent” to the NRC in violation of §§ 1001 and 2. The allegedly false material statements were:
1. “[i]n 1996 during 10 RFO, 100% of nozzles were inspected by visual examination” ...;
2. “[s]ince the [RFO10] video was void of head orientation narration, each specific nozzle view could not be correlated by nozzle number” ...;
3. “[t]he following pictures are representative of the head in the Spring 1996 Outage. The head was relatively clean and afforded a generally good inspection” ...;
4. “[b]ecause of its location on the head, [a pile of boric acid] could not be removed by mechanical cleaning but was verified to not be active or wet and therefore did not pose a threat to the head from a corrosion standpoint,” whereas, as the defendants then well knew, no action had been taken in 1996 to verify whether the boric acid was active or wet and, thus, not a corrosion threat;
5. “these attached pictures are representative of the condition of the drives and the heads” during the inspection during [RFO11] ...;
6. “[t]he photo for No. 19 depicts in the background the extent of boron buildup on the head and is the reason no credit is taken for being able to visually inspect the remainder of the drives,” whereas, as the defendants then well knew, other images from the 2000 inspection showed that the extent of boron buildup on the head was much greater than what was depicted in the photo of nozzle number 19.
We must uphold the conviction on this count if we conclude that there was sufficient evidence for a jury to find any one of these six allegations. See Dedman, 527 F.3d at 598. The evidence is strongest with respect to the first, fourth, fifth, and sixth statements, and so we limit our review to those allegations.
SL 2744, for which Geisen wrote the captions to the “representative” photographs, is the most direct evidence of Geisen’s participation in representing to the NRC that it was safe to continue operating the reactor because Davis-Besse had conducted adequate and thorough inspections in 1996, 1998, and 2000. Unlike in the previous serial letters, which Geisen did not directly draft, the captions were his own work product, and he testified that he did not consult Siemaszko, Goyal, or any other individual directly involved in the past inspections while drafting the captions. There is no question, therefore, that the alleged statements were made by Geisen and caused to be submitted to the NRC by him. In testimony, Geisen also admitted that the statements were false, although he denied that he knew that at the time that he wrote them. Therefore, the only element of § 1001 that remains is whether Geisen submitted those statements knowing that they were false.
Geisen attempts to shift blame for the misrepresentations and false statements in SL 2744 onto Siemaszko, stating that he *492had told Siemaszko to collect “representative” photographs and that he had based the captions on information from conversations with Siemaszko that had taken place previously but not specifically relating to the photographs. The government argues that Geisen “created” the captions by “interpreting] images and ma[king] critical representations about past inspections ... without any confirmed basis for doing so.”
Evidence presented at trial suggests that Geisen did task Siemaszko with selecting representative photographs of the inspections. It is undisputed that the photographs were not representative and showed far less boric acid buildup than existed. Given testimony presented at trial that Geisen had reviewed the inspection tapes in August 2001, the June 2001 memorandum, and the 2000 PCAQ in which Siemaszko had put a hold on the reactor due to the considerable deposits, a rational jury could find that Geisen knew of the greater extent of the boric acid deposits and the limited scope of the inspections by the time that SL 2744 was submitted on October 30, 2001. If Geisen had seen the prior inspection videos or read the reports, he would have known that the photographs were not representative of either the prior inspections or the current condition of the RPV head. A rational juror could therefore have found that Geisen’s captions indicating that the photographs were “representative” of the state of the RPV head were knowingly false and misleading.
Geisen’s attempt to blame Siemaszko for the content of the captions is also unavailing. Geisen testified that he wrote the captions based on past conversations with Siemaszko not related to the captioning, that he did not ask Siemaszko or others involved personally in the inspections to help with drafting the captions, and that he had not personally viewed the video tapes before writing the captions. The government also- argues that the statement that the deposits were “verified” as not wet or active was unconfirmed because Geisen did not ask anyone about the state of the deposits and did not have firsthand knowledge. Geisen testified that this caption “was based upon a conversation I had had with [Siemaszko] that he was reflecting back on a conversation he had with somebody else.” In prior statements to investigators, Geisen also attempted to blame Chimahusky for providing the information for the same caption. Chimahusky testified that he did not recall being asked about the 1996 inspections, and Goyal testified that he was not asked to help with this caption.
Additionally, as discussed previously with respect to count 3, evidence suggests that Geisen understood by October 30, 2001, that the NRC was very concerned with the thoroughness of the 1996 inspection and that his superiors wanted the plant to remain in operation until into the spring of 2002. Given this evidence of knowledge and motive to deceive and undisputed evidence that the photographs and captions were misleading, a rational jury could find that Geisen submitted SL 2744 to the NRC knowing that it contained false and misleading statements. Therefore, a rational jury could find that there was sufficient evidence to support a conviction of violating §§ 1001 and 2 on this count.
C. Count 1 — Concealing Material Facts
Count 1 of the indictment charged Geisen with “knowingly and willfully concealing] and covering] up, and causing] to be concealed and covered up, by tricks, schemes and devices, material facts in a matter within the jurisdiction of the [NRC], to wit, the condition of Davis-Besse’s [RPV] head, and the nature and findings of previous inspections of the *493[RPV] head” in violation of §§ 1001 and 2. The indictment specified ten allegations of concealment of a material fact:
1. causing SL 2731 to be forwarded to the NRC knowing that it
a. “deliberately omitted critical facts concerning the inspections and limitations on accessibility” and
b. “falsely stated that the inspections complied with ... Davis Besse’s [BACCP];”
2. falsely stating during an October 3, 2001, conference call with the NRC that a “100% inspection” of the RPV head with the exception of some areas took place in 2000;
3. representing the false fact that “[a]ll CRDM penetrations were verified to be free from ‘popcorn’ type deposits using video recordings from 11 RFO or 12RFO” at a meeting with the NRC on October 11, 2001;
4. causing SL 2735 to be forwarded to the NRC, which falsely represented that the entire RPV head had been inspected in 1996;
5. making false representations about the scope of the 1996 inspection, that “[a]ll CRDM penetrations were verified to be free from ‘popcorn’ type born deposits using video recordings from 10RFO, 11 RFO or 12RFO,” and that videos or eyewitness accounts confirmed this at a meeting with NRC staff on October 24, 2001;
6. causing SL 2741 to be forwarded to the NRC repeating false statements in SLs 2731 and 2735;
7. causing SL 2744 to be forwarded to the NRC with photographs falsely represented as “representative” of the condition of the RPV head;
8. causing SL 2745 to be forwarded to the NRC repeating false statements from SLs 2735 and 2741;
9. giving a presentation to the NRC with false information from SLs 2735 and 2741 to argue that the plant should stay open until RF013; and
10. giving a presentation to the FEN-OC Company Nuclear Review Board that falsely represented that a qualified visual inspection was performed in 1996 on all but four nozzles.
We need not analyze all ten allegations in the indictment individually because we must uphold Geisen’s conviction on this count if we find that there was sufficient evidence for a jury to find any one of the ten assertions. See Dedman, 527 F.3d at 598. We note first that because we have already found that there was sufficient evidence for a rational jury to find that Geisen caused SLs 2741 and 2744 to be forwarded to the NRC, knowing them to contain false statements and with intent to deceive, there was sufficient evidence to support a conviction on count 1 based on the sixth and seventh allegations.
However, even without relying on those statements, the government presented sufficient evidence to sustain this conviction. After the NRC’s negative response to SL 2731, Geisen met with NRC staff on at least five occasions — on October 3, 11 and 24 and November 8 and 28, 2001 — either via conference call or in person. The purpose of the meetings and conference calls was to reassure the NRC and to provide further information demonstrating the thoroughness of previous inspections.
During the October 3 call, Geisen stated that there had been a 100-percent inspection of the head barring five or six nozzles and, at an October 11 meeting, a slide authored by Geisen asserted that video recordings from RFO 11 and RFO 12 showed that the nozzles were free of boric acid. Geisen admitted that he realized that there was not a complete visualization from RFOll and RFO 12 directly after the *494meeting — and thus before submission of SLs 2735, 2741, and 2744 — but decided to correct the error later. It was at this juncture that the 1996 inspection became critical in order to represent that the nozzles had been fully inspected as recently as 1996.
The circumstantial and direct evidence discussed in our analysis of counts 3 and 4 with respect to Geisen’s knowledge of the state of the RPY head and the inspections could convince a rational jury that he knew that these representations were false by October 3 and 11. For similar reasons, a rational jury could find that Geisen knew that the information that he presented to the NRC at later meetings via slides authored by him knowingly included false and misleading statements. For example, a slide presented on October 24, 2001, stated that “the inspection results afford us assurance that all but 4 nozzle penetrations were inspected in 1996” and that “no penetration leakage was identified.” A jury, therefore, could convict Geisen on count 1 based on the misleading statements he made in meetings and calls with the NRC.
Geisen also argues that he bears no criminal responsibility for SL 2731 because he was not involved in its drafting and his only action was to sign off after the chain of review indicated on the green sheets was complete. It is undisputed that he signed the green sheet for SL 2731 on behalf of himself on August 28 and on behalf of his supervisor on August 30. Geisen admitted that his responsibility with respect to SL 2731 was to review the document. He testified that, in doing so, “[he] would have gone through the document looking for those pertinent sections that deal with the design of the plant and make sure that they sounded right to [him] as well as verify that the appropriate people from [his] staff were involved with the reviews and signed off on it.”
There is also both direct and circumstantial evidence that Geisen was aware of the limitations of previous inspections such that a rational jury could infer that he knew that the representations regarding those inspections in SL 2731 were false or misleading. For example, in the PCAQ for the the 2000 inspection, Siemaszko put the reactor on a restraint until the boron deposits were removed. Geisen added a page to the PCAQ removing the restraint because the RPV head was due to be cleaned, although it never was. In June 2001, Geisen signed off on a memorandum prepared by Goyal that stated that the plant was safe to operate until RF013 but that considerable boron deposits had impeded inspection of the nozzles in RF012. As part of that review, in light of Oconee, Goyal sent Geisen emails indicating that head cleaning should be a priority and that the center nozzles at Davis-Besse were of the same type as those that cracked at Oconee. While FENOC was drafting SL 2731, on August 11, 2001, Goyal sent another email to Geisen and to others in which he stated: “I indicated tha[t] we plan for 100% volumetric examination even if we do not commit to NRC.... It was pointed out that we can not [sic] clean our head thru the mouse holes and Andrew Seimaszko [sic] is requesting 3 large holes be cut in the Service Structure for viewing and cleaning.” This suggests not only that Geisen was involved in making sure that SL 2731 was sent to the NRC, but also that he was aware of the considerable impediments to previous inspections and that the inspections had not been conducted “in accordance with” the BACCP, as stated in SL 2731.
In conclusion, from this evidence, inter alia, a rational juror could find that Geisen knew that statements — which he re*495viewed — in SL 2781 were false and concealed the extent of the limitations to previous inspections and that he permitted those material statements to be sent to the NRC as such. We therefore affirm Geisen’s conviction on count 1.
TV. Exclusion of Evidence Claim
We review a district court’s decision to exclude evidence for abuse of discretion. United States v. Davis, 490 F.3d 541, 546 (6th Cir.2007). Furthermore, “[we] review[ ] de novo the [district] court’s conclusions of law and review[ ] for clear error the court’s factual determinations that underpin its legal conclusions.” United States v. Jenkins, 345 F.3d 928, 935 (6th Cir.2003) (citations omitted). We have found that these two standards of review are not in conflict, as “ ‘it is an abuse of discretion to make errors of law or clear errors of factual determination’ ” in evidentiary rulings. United States v. Baker, 458 F.3d 513, 517 (6th Cir.2006) (quoting United States v. McDaniel, 398 F.3d 540, 544 (6th Cir.2005)); see also United States v. Ganier, 468 F.3d 920, 925 (6th Cir.2006) (affirming the standard of review quoted in Baker).
“[T]he Constitution guarantees criminal defendants a meaningful opportunity to present a complete defense.” Crane v. Kentucky, 476 U.S. 683, 690, 106 S.Ct. 2142, 90 L.Ed.2d 636 (1986) (internal quotation marks and citation omitted). Although this guarantee includes the right “to present relevant evidence,” that evidence is subject to “reasonable restrictions” and must “bow to accommodate other legitimate interests in the criminal trial process.” United States v. Scheffer, 523 U.S. 303, 308, 118 S.Ct. 1261, 140 L.Ed.2d 413 (1998) (internal quotation marks and citation omitted). The Federal Rules of Evidence, including Federal Rule of Evidence 403, are such reasonable restrictions. See Varner v. Stovall, 500 F.3d 491, 499 (6th Cir.2007). Furthermore, a district court enjoys “ ‘wide discretion in determining the admissibility of evidence under the Federal Rules ... ’ [and t]his is particularly true with respect to Rule 403.” Sprint/United Mgt. Co. v. Mendelsohn, 552 U.S. 379, 384, 128 S.Ct. 1140, 170 L.Ed.2d 1 (2008) (quoting United States v. Abel, 469 U.S. 45, 54, 105 S.Ct. 465, 83 L.Ed.2d 450 (1984)).
Geisen argues that the district court improperly denied his motion to enter into evidence his rejection of a preindictment DPA. Geisen maintains that the DPA was probative of his state of mind because an innocent person is more likely to reject a DPA than is a guilty one. He further asserts that the rejection was necessary to counter impeachment evidence offered by the government regarding prosecution agreements offered to four testifying witnesses.
The district court addressed these arguments twice, once during pretrial conference, at which point the court tentatively expressed its intent to deny Geisen’s motion, and again during trial proceedings, when the court denied the motion in limine. In denying the motion, the district court first dismissed Geisen’s contention that the impeachment evidence of other witnesses necessitated inclusion of Geisen’s rejection of the same DPA because such impeachment evidence is routine “for the purpose of disclosing it to the jury so that they can judge whether the testimony is in exchange for the offer and acceptance by the government and the witness, not for the purpose of denial of guilt.” The district court went on to find that “there are more reasons to keep it out than to permit it to come in.” While acknowledging that the evidence may have some probative value, the district court noted that the jury’s weighing of the DPA was more complicat*496ed than the weighing of guilt or innocence because, as other witnesses testifying about offered DPAs admitted, accepting a DPA and the attendant statement of facts would affect an individual’s “viability to be employed within the nuclear industry.” Furthermore, the district court found that admitting the DPA would “open the door to cross examination on what he was told by his counsel; ... what he understood a[DPA] to mean for him, ... including his position for future employment in the nuclear industry and other employment[, and] his perception and maybe the discussion with counsel about ... [the weakness of the government’s case].” The district court concluded that “[t]here are just too many variables other than the explanation which would be permitted to the defendants on closing argument that that represented his denial of guilt.”
Geisen rests his argument heavily on a Second Circuit case holding that evidence of a rejection of an immunity offer is relevant to a defendant’s innocent state of mind. United States v. Biaggi, 909 F.2d 662, 690-91 (2d Cir.1990); see also United States v. Maloof, 205 F.3d 819, 824 (5th Cir.2000) (following Biaggi). The Biaggi court held that evidence of such a rejection is admissible if not otherwise “outweighed by the danger of unfair prejudice, confusion, or delay” under Rule 403. 909 F.2d at 691. The Eighth Circuit, however, has declined to adopt Biaggi in a case in which the defendant sought to introduce evidence of a rejection of a plea agreement. United States v. Greene, 995 F.2d 793, 798 (8th Cir.1993). The Greene court found controlling the reasoning of United States v. Verdoorn, 528 F.2d 103 (8th Cir.1976), which held that government proposals concerning pleas are inadmissible based on the rationale of Federal Rule of Evidence 408. Id. (noting that Rule 408 “relates to the general inadmissibility of compromises and offers to compromise” (quoting Verdoom, 528 F.2d at 107)). The Eighth Circuit found that there was “[no] relevant distinction between plea agreements and immunity agreements except, perhaps, as to the weight that jurors might give to them” and that “all the defendant is offering is a prior statement consistent with his plea of not guilty[, which is] hearsay, except in narrow circumstances.” 4 Id. (citing Fed.R.Evid. 801(d)(1)(B)). As the government points out, “mak[ing] evidentiary use of [Geisen’s] rejection of the offer — as opposed to the offer itself — [would be] difficult (if not impossible) to entangle.”
We have not previously addressed this question, and find no reason to reach it now because, on the facts of this case, the exclusion of evidence of Geisen’s rejection of the DPA did not constitute an abuse of discretion by the district judge necessitating reversal even under Biaggi. The Biaggi court held that “the probative force of a rejected immunity offer is clearly strong enough to render it relevant” because an immunity offer would “preclude all exposure to a conviction and its consequences.” 909 F.2d at 691-92 (emphasis added). A deferred prosecution agreement, however, does not foreclose all exposure to a conviction and its consequences in the same way, especially given the implications for Geisen’s employment. Refusing to accept the DPA, therefore, is not as probative of a “consciousness of innocence” as the immunity offer at issue in Biaggi. The Biaggi court itself asserted that a plea agreement would be less probative and declined to *497reach whether a district court would be required to admit evidence of refusal of such an agreement. Id. at 691.
Furthermore, and more importantly for the case before us, the Biaggi court acknowledged that “[i]t is a closer question whether the District Judge exceeded her discretion under [Rule 403] to bar relevant evidence [of the immunity offer] whose probative value is outweighed by the danger of unfair prejudice, confusion, or delay.” Id. The Biaggi court “recognize[d] the latitude of a district judge in making Rule 403 determinations” but found that the district judge had based her Rule 403 determination on the erroneous assumption that the immunity offer was not at all relevant. Id. In the instant case, however, the district court acknowledged that Geiseris rejection of the DPA may be probative of an innocent state of mind, but found the probative value outweighed by the considerable avenues of inquiry that would be opened by admitting the evidence. The court also expressed concern that prejudice to the government would result because “too many variables [existed] other than the explanation which would be permitted to the defendants on closing argument that that represented his denial of guilt” and that much relevant testimony would be privileged. The district court’s reasoning for denying the motion in limine, therefore, does not demonstrate an abuse of discretion in the same manner as it did in Biaggi. We therefore find no error in the district court’s exclusion of evidence of Geisen’s rejection of the DPA.
V.
For the foregoing reasons, we affirm Geisen’s conviction on counts 1, 3, and 4 of the indictment.
. Cook was indicted on all counts except count 4.
. We decided Siemaszko's appeal this day in a separate opinion. See United States v. Siemaszko, No. 09-3167, 2009 WL 1856609 (6th Cir.2010).
. In distinguishing his case from Mari, Geisen argues that Mari misapplied Griffin and relies on cautionary language from this and other circuits regarding the use of the deliberate ignorance instruction. He also points to the district court’s hesitance to determine whether such an instruction was appropriate in this case as evidence that the instruction was improper. We are not persuaded. Mari remains controlling law in this circuit, and the district court’s observation that this was a close issue does not necessarily render his ultimate determination of the issue arbitrary or capricious.
. The Eighth Circuit also noted, however, that because the defendant did not intend to testify, there would be no opportunity to cross-examine him regarding the agreement. This concern is nol present in this case, because Geisen did take the stand and could have been cross-examined.