ORDER
HONORABLE STEVE C. JONES, UNITED STATES DISTRICT JUDGEIn this Employee Retirement Income Security Act (ERISA) case, both parties move for judgment on the administrative *1384record. Docs. 21 & 25.1 Because Greater Georgia Life Insurance Co. (GGL) abused its discretion in denying benefits, its motion is DENIED. Because that decision also was de novo wrong, Ricky Madison’s motion is GRANTED,
I. BACKGROUND
Ricky Madison previously worked for Mansfield Oil Co. as a transport driver “responsible for the loading, transporting, and delivery of fuel.” Doc. 26-1 at 4. That job required him to sit “for extended periods of time,” lift and transport “moderately heavy objects,” “climb in and out of a truck,” and depress the truck’s clutch. Id. at 5. Problems with his left knee made those activities, and thus work, unbearable beginning on February 24, 2014. Id. at 2.
A. Course of Treatment
The next day, Madison saw an internist, Dr. Macolm Simpson, about his knee pain. Doc. 26-1 at 65. A visit to another internist, Dr. Donald Hanchett, followed. Id. at 69. After observing “loose bodies in the joint space of the left knee,” and “mild medial compartment osteoarthritis,” Han-chett referred Madison to an orthopaedic specialist, Dr. Monique Grey-McBride. Id. at 74.
She ordered an MRI, which revealed “evidence of osteochondral defect,”2 “[t]wo large osteocartilaginous loose bodies,” and a possible meniscus tear. Doc. 26-1 at 81. A physical exam revealed reduced range of motion in Madison’s left knee (15 degrees extension, 110 degrees flexion,3 as eom-pared to 0 deg. Extension and 125 deg. flexion in his right knee), and reduced quadriceps strength in the left leg (3/5). Id, at 80. McBride concluded that he would “likely benefit from knee arthroscopy,” but that “[t]he initial osteochondral injury is 20+ years old and is not amenable to repair.” Id. at 82. She instructed Madison to limit the “frequency, duration, and intensity of any activity that aggravates the pain until symptoms subside.” Id.
Another orthopaedist, Dr. William Craven, performed the recommended procedure on April 7, 2014. Doc. 26-1 at 102. He found “three extremely large loose bodies ... which had been preventing [Madison] from fully extending and flexing his knee.” Id. Craven also “incised and excised” the meniscus tear, and “did a chondroplasty on the trochlea.” Id. Overall, Craven saw “significant pathology” in Madison’s knee. Id.
Nine days later, Madison’s range of motion and knee strength, unsurprisingly, remained limited (80 or 90 deg. flexion and 10 deg. extension in the left, 110 and 0 deg. respectively in the right, with 4/5 strength across the board in the left). Doc. 26-1 at 98. He described his pain as “8/10,” but was “able to put on socks,” “get in/out of a car,” bathe, and shop without assistance. Id. at 97. Craven referred Madison to physical therapy and gave him a steroid prescription. Id. at 99.
One month later, Madison followed up with Craven again. Doc. 26-1 at 126. Madison reported 5/10 pain and the same activities of daily living. Id. He once again (and *1385in spite of participating in physical therapy, see id. at 128) displayed reduced range of motion and strength as well (100 deg. flexion and 10 deg. extension in the left, 120 and 0 deg. respectively in the right, with 4/5 strength across the board in the left). Id. at 129. Craven ultimately provided a work status report a month after that and placed Madison “on modified activity at work and at home from 5/16/2014 through 7/16/2014.” Doc. 26-1 at 112. That meant he could stand and walk up to 25% of a given shift, and never squat, kneel, bend his knees, or climb ladders. Id. at 112. Those restrictions meant Madison could not perform his fuel truck driver job.
A July 8, 2014 Craven visit revealed more of the same. The knee continued to cause Madison problems, including swelling, pain upon standing longer than ten minutes, and reduced range of motion and strength (100 deg. flexion and 10 deg. extension in the left, 125 and 5 deg. respectively in the right, with 4/5 strength across the board in the left). Doc. 26-1 at 140-41. Craven provided a steroid injection in the knee because of “recurrent episodes of swelling and pain” and instructed Madison to return for a check-up in six weeks. Id. at 141.
At that follow-up appointment on August 20, 2014, Craven observed tenderness and a slightly reduced range of motion and strength (120 deg. flexion and 10 deg. extension in the left, 125 and 0 deg. respectively in the right, with 4/5 strength across the board in the left). Doc. 26-1 at 146. He ordered a second MRI,4 diagnosed Madison with osteoarthritis, and noted:
More subjective than objective findings. Patient requesting total Disability from job, explained that he did not qualify for total disability. Recommended that if he had a question about disability he should obtain an functional capacity evaluation test. Submitted sedentary work status.
Id. at 147.
A trip to the neurologist was next. Dr. Ramesh Kumar found no neurological cause of Madison’s pain and confirmed that his “symptoms are of left knee origin.” Doc. 26-1 at 171. With no treatments to offer, the doctor recommend an ortho-paedic follow-up. Id
Back to Craven Madison went. Despite multiple previous visits, the nurse identified the November 5, 2014 appointment as an “initial evaluation of [Madison’s] knee concerns.” Doc. 26-1 at 166. Craven found “no swelling” and “no effusion,” and negative Lachman, Drawer, and Pivot Shift tests. Id. at 167. Madison had symmetrical range of motion and strength in both knees (125 deg. flexion, 0 deg. extension, 5/5 strength) and symmetrical quads (i.e., no atrophy in his left leg). Id. In light of those measurements, Craven observed that “[t]he left knee appears normal, there is no swelling/effusion. No instability, Post op xrays and repeat MRI show no orthopedic lesion to explain his complaints.” Id, He concluded that Madison “ha[d] no orthopedic problem to treat” and released him from his care. Id. at 168. In a work status report issued the same day, Craven deemed Madison “able to return to work at full capacity on 11/11/2014.” Id. at 157.
Madison decided to get a second opinion and went to see Dr. Sangmin Shin on November 11, 2014. Doc. 26-1 at 160. Shin observed no swelling, but did note effusion’s presence, as well as quadricep atro*1386phy. Id. at 161. He examined Madison’s MRI results, and, like every other doctor, found a “chondral defect over trochlea.” Id. Indeed, Shin stated that Madison is:
symptomatic from his trochlea chondral defect. He has been deconditioned since the surgery. He had great relief from steroid injection previously therefore I offered him steroid injection today, which he elected to receive. I advised him that he needs to start on Quad strengthening regimen. I gave him 1 month work note today for him to recover from deconditioning. He understands that this will be only work note that I can provide for him. I also gave him hinged knee brace to be worn for ambulation only. We’ll check back with him in a month. Patient understands and agrees with assessment and plan.
Id. at 162. Shin also prescribed an exercise routine and, like Craven, diagnosed osteoarthritis. Id.
Christmas Eve 2014, Madison sought yet another opinion, seeing Dr. Michael Bakheet’s (another orthopaedic surgeon) physician’s assistant, Amanda Appling. Doc. 26-1 at 186. Her exam revealed “[tjrace effusion,” some tenderness, and noted that Madison was “unable to actively fully extend the left knee,” though Appling could “fully extend” it. Id. at 186-87. Left knee flexion was “100 degrees but he has assist with this motion. 4/5 strength noted on quad flexion and extension against resistance.” Id. at 187. After the exam, Ap-pling stated that “IT band tightness and friction ... can be a cause of [Madison’s] lateral knee pain,” though she did “not know the exact cause of’ that pain and his “inability to walk normally.” Id. She recommended that he “get back into physical therapy for modalities, strengthening, and RANGE OF MOTION.” Id. at 189. Ap-pling also prescribed a topical lidocaine ointment to help with pain. Jd. at 188.
After being told by “orthopedics” that “there is nothing more that can be done for him,” Madison saw Dr. Lewin Chua-chiaco, an internist, on January 5, 2015. Doc. 26-1 at 239. Madison reported that he “scheduled more physical therapy,” but wasn’t “sure whether” it would help, particularly since home-based therapy, walking, and stretching did not improve his symptoms. Id. at 239-40. Chuachiaco’s musculoskeletal exam showed “discomfort on flexion and extension of the left knee,” “some swelling,” “atrophy of the left quadriceps muscle,” and “[s]light discomfort on the back of the left knee.” Id. at 241. He thereafter ordered another MRI,5 inferred Madison to orthopaedics, and ordered a chart review. Doc. 26-2 at 1. As did Appling, Chuachiaco also implored Madison to “get the therapy done.” Id.
Apparently in response to Chuachiaco’s chart review order, Craven entered a progress note dated 1/29/2015 under the January 5, 2016 encounter date. Doc. 26-2 at 2. It stated simply: “Reviewed. Agree with most recent eval. Patient does not need outside evaluation. He has had 3 evaluation^] that all agree.” Id. at 2.
Rewind two days and on January 27, 2015, Madison returned to his original or-thopaedist, Dr. McBride, per Chuachiaco’s referral. Doc. 26-2 at 1, 3. She reviewed his recent MRI and conducted a physical exam, which showed reduced range of motion and strength (8 deg. extension, 105 deg. flexion, 4/5 quad strength). Id. at 6. Based on those findings and the imaging results, McBride concluded that Madison’s pain was “most likely [due] to symptomatic *1387chronic osteochondral defect/chondromala-cia of lateral femoral condyle and weakness.” She ordered a round of Supartz injections and recommended that Madison “[pjursue physical therapy given decondi-tioning of knee,” noting that “it was ordered for him on 12/24/2014.” Id. at 7.
Three days later, on January 30, 2015, Madison went back to see Chuachiaco. He reported continued problems standing and stated that he did “not know when he [would] start the physical therapy.” Doc. 26-2 at 10, An exam revealed “some swelling over the left knee,” “some difficulty with weight bearing,” and “[g]ood passive range of motion,” but continued “pain and discomfort on flexion and extension.” Id. at 12. Chuachiaco’s “plan” involved awaiting the Supartz results and physical therapy. Id. at 13. He gave Madison a six week “stay off work” note. Id. ■
McBride administered the last two Su-partz injections on February 24 and March 3, 2015. Doc. 26-2 at 21, 26. Madison reported some improvement, but his osteoarthritis and chondromalacia diagnoses remained. Id. Not two weeks later, Madison returned to Chuachiaco because two days after the last Supartz injection his knee pain “got so intense that he stayed up all night.” Id. at 29.
Madison reported that financial reasons prevented him from scheduling therapy, but that he “was [also] concerned about the physical therapy making the knee worse.” Doc. 26-2 at 29. Chuachiaco, after observing knee swelling and tenderness, advised Madison that he nevertheless needed supervised therapy, discontinued ibuprofen, and prescribed two other pain relievers, including a narcotic. Id at 31-32.
Madison, on yet another visit to Chua-ehiaco, reported on April 2, 2015 that he finally scheduled physical therapy. Doc. 26-2 at 34. Chuachiaco noted no swelling at that visit, but did observe a decreased range of motion and discomfort. Id. at 36-37. He ordered continued therapy, medications, and a follow-up in six weeks. Id at 37.
At that visit, on May 14, 2015, Madison stated that “he ha[d] tried physical therapy as per recommendations but the knee pain got so excruciating after 2 sessions [that] he had to stop.” Doc. 26-2 at 218. His subjective reporting by that point reflected the grimmest self-assessment to date. See, e.g., id. at 219 (“He says that he cannot stand for more than 5 to ten minutes. He says that he can only stand long enough to brush his teeth. In his opinion the knee is worse than when I saw him six weeks ago. In fact, he says that it is almost as bad as it was before he had the surgery.”). Chuachiaco’s exam noted swelling and a “limited range of motion,” with Madison only able “to extend his knee to 20 degrees,” and “able to flex ... to about 90 degrees but only with some effort and also slowly,” while grimacing the entire time. Id. at 221, He extended Madison’s work excuse for another eight weeks. Id. at 222.
After his lawyer convinced defendant Greater Georgia Life Insurance Co. (GGL) to allow a second appeal of denied long-term disability benefits (see infra), Madison obtained a functional capacity evaluation (FCE) on August 6, 2015 (doc. 26-2 at 191-204), originally recommended by Dr. Craven almost a year earlier. Doc 26-1 at 146; Administered by Daniel Navarro, a licensed physical therapist, the FCE concluded, among- other things, that Madison is “unable to safely or effectively drive a tanker truck that requires the operation of a clutch with his left leg.” Doc. 26-2 at 192. It found his floor-to-waist lifting ability limited to ten pounds, and recommended avoiding any kneeling, squatting, balancing, or ladder, climbing. Id.
*1388Navarro also measured left quad atrophy (2.25 centimeters smaller than his right quad), knee strength (4-/5), and range of motion (124 deg. flexion, -5 deg extension), all of which pointed to impairment. Doc. 26-2 at 198. And to counterbalance any malingering, the FCE included a number of measures designed to ferret out lack of effort. None pointed to Madison short-changing the exercises.6
B. Disability Benefits Award, Denial, and Appeal(s)
GGL, Mansfield’s disability insurance provider, began paying Madison long-term disability benefits on May 28, 2014 (this, after Madison exhausted short-term benefits through GGL), about seven weeks after his knee surgery. Doc. 11 at 9. On November 19, 2014, GGL notified Madison that it was terminating benefits based on Craven’s November 5, 2014 progress note. Doc. 26-2 at 97. After it had a chance to review Shin’s progress notes, however, GGL extended benefits through December 15, 2014. Id. at 98.
Pursuant to its ongoing evaluation of Madison’s eligibility, GGL initially denied benefits on January 27, 2015, asserting that he lacked a disabling condition after August 20, 2014 and that since benefits had been paid through December 15, 2014, it owed him nothing further. Doc. 26-2 at 105, 108. Before doing so, it engaged a third-party to review Madison’s medical records.
Dr. Vicki Kalen, a board-certified ortho-paedic surgeon, performed the review on January 7, 2015 and concluded, based solely on GGL’s “referral form and submitted clinical highlights” (doc. 26-2 at 48), that no physical condition supported by clinical evidence functionally impaired Madison. Id. at 49. She acknowledged Bakheet’s December 24, 2014 weakness and range of motion findings, but discounted its credibility based on Craven’s November 5, 2014 exam that showed full strength and normal range of motion, and his August 20, 2014 bilateral atrophy measurements, which showed “no significant difference.” Id. “Based on the examination, x-ray and MRI of late 08/2014,” Kalen concluded, “there is no documentation of impairment that would have precluded a return to work at his usual occupational level at that time (08/2014).” Id. For whatever reason (perhaps the clinical highlights omitted them), Kalen never referenced Shin’s November 2014 exam, Chuachiaco’s January 2015 exam, or the August 20 measurements, which actually showed one centimeter of left leg atrophy just above the knee. Doc. 26-1 at 147.
Madison appealed the benefits denial on February 17, 2015. Doc. 26-2 at 103. In his appeal letter, Madison specifically mentioned his January 27, 2015 McBride visit and the resulting six week work excuse. Id.
GGL actually received a copy of McBride’s progress notes from that visit as well as notes from the January 5, 2016 Chuachiaco visit (including Craven’s January 29, 2015 note responding to Chuachia-co’s “chart review” request), on either April 3, 2015 (doc. 26-2 at 125), or April *138914, 2015. Doc. 26-3 at 123. GGL then hired a second outside reviewer, Dr. Daniel Benson, an orthopaedic surgeon, to examine Madison’s medical records. Doc. 26-2 at 67 (Benson report dated April 22, 2015). Benson’s report reflects that he reviewed records dating from February 2014 through Madison’s April 2, 2015 Chuachiaco visit, not merely a clinical summary. Conspicuously absent from the list of records, however, were Chuachiaco’s January 5, 2015 notes; McBride’s January 27, 2015 notes, Craven’s January 29, 2015 note, and Chua-chiaco’s January 30, 2015 note and work restriction. Id. at 71-72.
Without ever discussing those particular records (though he did mention virtually every other visit from February 2014 to March 2015), Benson concluded that Madison’s subjective complaints were inconsistent with the objective medical evidence. Doc. 26-2 at 69. In particular, he stated that post-op, Madison had a “a stable knee with no positive examination signs.” Id. “His strength ha[d] repeatedly been rated as 5/5,” noted Benson, and “[although seen by several providers after his full release to work on 11/11/2014, none provided a firm explanation for [Madison’s] continued self-directed complaints of pain.” Id. “[B]ased on the record with an emphasis on the final evaluation of ll/[5]/2014” with Dr. Craven, Benson concluded that “[t]here are no restrictions or limitations that would impede[Madison] from performing his occupation beyond 12/14/2014.” Id. at 70. Based largely on Benson’s review (see doc. 26-2 at 125), GGL denied Madison’s appeal on April 29, 2015. Id. at 127.
After Madison retained counsel, GGL agreed to perform a “courtesy review of [its] claim determination” and agreed to extend the time for Madison to submit additional medical records. Doc. 26-2 at 161. He did so on August 21, 2015, including with his appeal the FCE, Chuachiaco’s August assessment, and updated medical records. See, e.g., id. at 171-180. GGL hired a third reviewer, Dr. William Andrews, another orthopaedic surgeon, to examine those and other records. Id. at 73.
Andrews’ report indicates he examined records that began in February 25, 2014 and ran through August 2015, including the FCE and Chuachiaco visits in April, May, and July 2015. Doc. 26-2 at 77-78. Absent from the list of records are the January 27, 2015 McBride visit, Chuachia-co’s January 5, 2015 notes, and Chuachia-co’s January 30, 2015 note and work restriction. Id. at 78. Despite the FCE, which he never even mentioned in passing, Andrews concluded that “[t]here is no clinically relevant medical evidence to support [Madison’s] self-reported complaints.” Id. at 76. Despite the FCE, and per “Craven’s evaluation on 11/05/2014,” Andrews concluded that Madison possessed a “full range of motion, normal motor function and no quadriceps atrophy.” Id. And, despite the FCE, he “would place no restrictions on the claimant that would impede his ability to perform his job from 12/16/2014—present.” Id.
GGL denied the courtesy appeal on September 9, 2015 based solely on Andrews’ report and without mentioning the FCE. Doc. 26-3 at 44. Madison filed this action less than a month later. Doc. 1.
II. STANDARD OF REVIEW
Courts reviewing ERISA plan administrator decisions:
(1) Apply the de novo standard to determine whether the claim administrator’s benefits-denial decision is “wrong” (i.e., the court disagrees with the administrator’s decision); if it is not, then end the inquiry and affirm the decision.
(2) If the administrator’s decision in fact is “de novo wrong,” then determine whether he was vested with discretion in *1390reviewing claims; if not, end judicial inquiry and reverse the decision.
(3) If the administrator’s decision is “de novo wrong” and he was vested with discretion in reviewing claims, then determine whether “reasonable” grounds supported it (hence, review his decision under the more deferential arbitrary and capricious standard).
(4) If no reasonable grounds exist, then end the inquiry and reverse the administrator’s decision; if reasonable grounds do exist, then determine if he operated under a conflict of interest.
(5) If there is no conflict, then end the inquiry and affirm the decision.
(6) If there is a conflict, the conflict should merely be a factor for the court to take into account when determining whether an administrator’s decision was arbitrary and capricious.
Blankenship v. Metro. Life Ins. Co., 644 F.3d 1350, 1355 (11th Cir. 2011).
III. ANALYSIS
A. De novo Review
“To determine under the first step whether the Board’s decision was ‘wrong,’ the Court must examine the terms of the Plan and the administrative record to determine whether [it] agrees with the Board’s decision.” Grant v. Bert Bell/Pete Rozelle NFL Player Retirement Plan, No. 1:09-CV-1848, 2011 WL 6711595, at *5 (N.D.Ga. Dec. 16, 2011). Review “is limited to consideration of the material available to the administrator at the time it made its decision,” Blankenship, 644 F.3d at 1354, but that decision receives no deference. See Williams v. BellSouth Telecomms., Inc., 373 F.3d 1132, 1137-38 (11th Cir. 2004), overruled on other grounds by Doyle v. Liberty Life Assurance Co. of Boston, 542 F.3d 1352 (11th Cir. 2008).
1. Records Before the Court
As a threshold matter, the parties dispute what GGL decision should be reviewed and thus what records the Court may properly consider. Madison sees GGL’s September 2015 courtesy review denial as the relevant landmark, and thus argues for consideration of all records. Doc. 32 at 2. GGL, on the other hand, insists that it issued a final decision denying benefits on April 27, 2015, and thus only records received before then are properly before the Court. Doc. 28 at 8.
As noted, review is “based upon the facts as known to the administrator at the time the decision was made.” Jett v. Blue Cross & Blue Shield of Ala., Inc., 890 F.2d 1137, 1139 (11th Cir. 1989). Hence, “once a final denial of an ERISA claim has occurred and the claims administrator has advised the plaintiff that she has exhausted her administrative remedies, a plaintiffs subsequent submissions in support of her finally-determined benefits claim, even if voluntarily accepted by the administrator, do not expand the scope of the record for determining ERISA liability. Instead, those records are meaningless non-events insofar as an administrator’s exposure to ERISA liability is concerned.” Blair v. Metro. Life Ins. Co., 167 F.Supp.3d 1272, 1278 (N.D. Ala. 2016).
In Blair, the plaintiff, like Madison, asked MetLife for a courtesy review of her claim after it issued a final appeal decision. Blair v. Metro. Life Ins. Co., 955 F.Supp.2d 1229, 1234 (N.D. Ala. 2013).7 *1391The company agreed “to conduct one further review and ... [to] accept any additional information counsel wished [it] to consider ... until July 31, 2009.” Id. (alterations and quotes omitted, alterations added). For the next two years, Blair’s attorney proceeded to send letters that at times indicated he would submit yet more records, at others that he was “done submitting additional materials.” H. at 1235. Critically, before MetLife ever issued a decision on the courtesy review, Blair filed suit, arguing that MetLife violated the 90 day deadline for appeal decisions, denied her “full and fair review,” and thus that her new evidence should be reviewed de novo. Id. at 1236-37.
It is in that context that Blairs rule— “once a final denial of an ERISA claim has occurred and the claims administrator has advised the plaintiff that she has exhausted her administrative remedies, a plaintiffs subsequent submissions in support of her finally-determined benefits claim, even if voluntarily accepted by the administrator, do not expand the scope of the record for determining ERISA liability”—must be viewed. Blair’s courtesy review never resulted in a final decision that a court could review. Any records submitted after Met-Life’s previous denial (a true final decision) in connection with the never-completed voluntary review, therefore, were “meaningless non-events” for purposes of reviewing the earlier decision. Blair, 167 F.Supp.3d at 1278.
Here, by contrast, GGL issued a final decision on Madison’s requested second review. It had before it for consideration the FCE and other records submitted after its April 27, 2015 appeal denial (or at least it submitted those records to Andrews for his review, see 26-2 at 78). Unlike Blair, this Court has a finalized decision making process to review. What’s more, GGL acknowledged as much in its denial letter, where it informed Madison, for the first time, that he had “exhausted his rights for an administrative review under th[e] Plan,” and could thus file suit. Doc. 26-3 at 47.8 However legally inaccurate that may be (Madison ERISA-ex-hausted once GGL denied his first appeal, see doc. 26-2 at 130), it reinforces that GGL saw its courtesy review decision as giving rise to potential liability.
That GGL had no contractual or statutory obligation to conduct that review9 offers ERISA liability protection only for Blair-like claims, where an alleged violation stems from a voluntary review that never reaches decision: To hold otherwise, and untether Blair’s rule from a final decision guidepost—i.e., hold that voluntary reviews can never generate liability— incentivizes administrators to grant requests for courtesy review and then either slow-roll a decision or provide no meaningful review at all in an effort to delay suit. That’s not what ERISA requires and not what this Court permits. Indeed, ERISA contemplates “voluntary levels of appeal,” and in doing so never removes those levels from judicial review. See 29 C.F.R. § 2560.503-l(c).10
*1392Again, Madison, unlike Blair, waited until after GGL’s courtesy review denial to file suit. Put differently, and as Madison correctly notes, he “waited until GGL ... actually made [a] ‘final decision’” in the courtesy appeal. Doc. 32 at 5. Because that decision stemmed from a review that included the FCE and other post-April 30, 2015 records, this Court can consider them now.
2. GGL’s Decision
To ascertain whether GGL de novo erred by denying Madison benefits, the Court must first examine the LTD plan (“Plan”) terms. See Brannon v. BellSouth Telecomms. Inc., 318 Fed.Appx. 767, 769 (11th Cir. 2009) (“In our de novo review, we turn first to the plan itself. See 29 U.S.C. § 1104(a)(1)(D) (providing that an ERISA plan administrator must “discharge his duties with respect to a plan ... in accordance with the documents and instruments governing the plan”)”). GGL’s plan provided long-term disability benefits to injured Mansfield employees who were (1) “unable to do the Material and Substantial Duties of [their] Own Occupation;” (2) “receiving Regular Care from a Physician for that Injury or Illness;” and (3) whose “Disability Work Earnings, if any, are less than or equal to 80% of Your Indexed Monthly Earnings.” Doc. 26-3 at 181. GGL must receive “Proof of ... Disability,” which “means evidence satisfactory to [GGL] that the terms and provisions of the Policy have been met.” Id. at 172. Consequently, GGL was “de novo wrong” to deny benefits if the record shows that Madison submitted evidence that he could not perform the duties of a fuel truck driver for the time period in question.
As a general matter, drivers must load, transport, and deliver fuel. Doc. 26-1 at 4. For Madison, that meant an average of “6 trips a day,” some of which exceeded thirty minutes of continuous driving, during which he had to press a stiff clutch with his left leg. Doc. 26-2 at 233.11 “In addition to just driving the tanker truck, [Madison] had to be able to get in and out of the cab at least 48 times a day.” Id. That required navigating a two foot high step. Id. While not in the cab, Madison lifted, carried, placed, and replaced 70 pound fuel hoses, some of which he did while crouching. Id. at 451-52.
GGL insists that “[a]fter December 14, 2014, no doctor provided the medical documentation necessary to support a finding of disability under the Plan.” Doc. 25 at 19. In fact, it says, “seven board certified doctors ... found that Madison could return to work by December 15, 2014, or earlier.”12 Doc. 25 at 16. (emphasis in original). “In addition, none of the seven doctors explained Madison’s subjective complaints of pain. Dr. Craven found no objective evidence of disability; Dr. Kumar found no neurological explanation; Dr. Bakheet did not know the exact cause of [Madison’s] vague knee pain and tightness; and three independent reviewers found no clinically relevant medical evidence to support [Madison’s] self-reported complaints.” Id. at 17 (quotes and cite omitted). Faced with those facts, and because the GGL plan *1393required Madison to prove continuing disability, GGL insists it was de novo correct to deny benefits on January 27, 2015. Id. at 18.
Madison’s subsequent document submissions, contends GGL, failed to provide adequate proof of disability. Doc. 25 at 19 (“After December 14, 2014, no doctor provided the medical documentation necessary to support a finding of disability under the Plan.”). Chuachiaco, for example, never, according to GGL, stated “that Madison should refrain from working,” and instead referred him “to physical therapy.” Id. He never responded to the record reviewers’ attempts to contact him, and his August 19, 2015 letter, GGL notes, “did not ... state that Madison ... could not perform his job duties.” Id.
Even if tidbits of Chuachiaco’s notes amount to conflicting evidence, says GGL, that “scintilla of ... evidence does not make [its] decision de novo wrong.” Doc. 25 at 20. Neither he “nor the [FCE] states that Madison could not perform his job duties from December 14, 2014, to some later date. At most, both Dr. Chuachiaco’s note and the Functional Capacity Examination show that Madison had a less-than-100%-healthy left-knee.” Id.'
Indeed, the FCE “is hardly reliable evidence of disability.” Id. Such an exam lacks “a medical component” and the therapist has no “medical training to review the claimant’s medical history.” Id at 21. And, claimants can “fake” an FCE and render its results unreliable. Doc. 28 at 10 n.3. In any event, says GGL, Madison’s failure to attend doctor-ordered physical therapy explains the poor FCE results. Id. at 11. “Hence the evaluation—which was conducted almost eight months after medical doctors deemed Madison healthy (and sixteen months after his original injury)— cannot possibly show Madison’s continuing disability from December 14, 2014, until August 8, 2015, and beyond.” Doc. 25 at 22.
Madison disagrees. Pointing to his job duties and treating physicians’ “symptomatic chronic osteochondral defect” diagnosis, Madison insists the proper question is “to what extent do [that] condition and its symptoms impact [his] ability to perform his regular job duties?” Doc. 21-1 at 22. And that, says Madison, is best answered by an FCE. Id (citing Stiltz v. Metropolitan Life Ins. Co., No. 1:05-cv-3052-TWT, 2006 WL 2534406, at *12 (N.D. Ga. Aug. 30, 2006) (“An FCE is often considered the “best means of assessing an individual’s functional level.”)). Indeed, “[t]he FCE is the only measure of how Madison’s condition impacts his ability to lift and carry objects or squat or kneel to the ground.” Id. at 24.
The FCE testing here, Madison notes, unequivocally found him unable to perform his job’s requirements. Doc. 21-1 at 24. Unlike doctors’ exams, it actually measured his “ability to stand and flex/bend his knee in an attempt to squat to the ground under his own weight which is necessary for attaching and detaching fuel hoses and safe lifting techniques.” Id. at 25. Perhaps more importantly, it’s consistent with multiple doctors’ clinical evidence (i.e., range of motion, atrophy, and strength measurements). Id. Finally, notes Madison, the FCE contained several effort verification measures that prove he “gave it all he had” and did not try to game the exam. Id. at 24. Those veracity verifiers enhance the FCE’s objectivity and thus its evidentiary value such that GGL’s contrary decision— that failed even to take the FCE into account—is de novo wrong.
Were that not enough, Madison’s own, credible, statement, he contends, supports a disability finding and deserves weight (GGL gave it none). Doc. 21-1 at 27-28. GGL’s record reviewers, by contrast, de*1394serve no weight. Id. at 28-29. None ever examined Madison or considered Chua-chiaco and McBride’s 2015 exams and work restrictions (or Craven’s January 29, 2015 concurrence with “three previous” exams), two had no chance to consider the FCE, and Andrews, the only reviewer who had access to it, never mentioned the FCE in his report, instead inexplicably focusing almost entirely on Craven’s November 5, 2014 exam. Id. at 29. Faced with a credible, outcome-determinative FCE and consistent, supportive, medical records, GGL should have found Madison disabled—i.e., unable to perform the duties of his job as a fuel truck driver—under the Plan, That it did not, contends Madison, constitutes de novo error.
The Court agrees. From April 2014 until; GGL’s voluntary review denial in September 2015, Madison saw five different physicians at least sixteen times.13 Each time, whether the doctor was an orthopaedic surgeon or an internal medicine specialist, Madison reported and the doctor observed, symptoms like flexion and extension discomfort, swelling, and atrophy. See, e.g.,' doc. 26-1 at 239 (Madison’s January 5, 2015, pre-benefits denial, subjective report of pain and discomfort). And each time, objective measurements and/or observations (range of motion and strength) supported his subjective reporting. See, e.g., id. at 241 (Chuachiaco atrophy and discomfort observation).
So too did Madison’s three MRI scans, reveal musculoskeletal defects. The March 2014 pre-operative MRI showed “evidence of osteochondral defect.” Doc. 26-1 at 81. The August 2014 MRI showed a “full-thickness cartilage defect of the trochlea at the ... lateral femoral condyle.” Id. at 173. And the January 13, 2015 MRI revealed a “[cjhronic osteochondral defect of the anterolateral femoral condyle.” Id. at 204. No contrary imaging exists.
Most importantly, Madison’s FOE provides ample objective medical evidence that his injury rendered him, “unable to do the Material and Substantial Duties of [his] Own Occupation.” Doc. 26-3 at 181. Navarro, like every other doctor to examine Madison (excepting Craven on November 5, 2014, discussed infra), found atrophy in the left thigh (and calf), reduced range of motion, and reduced muscle strength. Doc. 26-2 at 198. Critically, those findings mirrored objective measurements taken by Shin, McBride, Craven .(other than his aberrant November 5, 2014 exam), and Chuachiaco.
Unlike those other exams, the FCE involved occupational abilities tests that measured Madison’s capacity for activities that resembled his job’s. For example, he “was asked to floor to waist lift” both ten and twenty pound weights. At ten, Madison displayed “safe control during each repetition,” but at twenty he “became unsteady on his feet ... due to his inability to have an equal base of support.” Id. at 201.
The same pattern repeated with a waist-to-overhead lift of the same ten and twenty pound weights. Doc. 26-2 at 201. And, when Madison carried two ten pound weights “for 30 feet over level concrete terrain,” he did so “inefficienctly and with significant difficulty.” Id. (noting that Mad*1395ison’s heart rate “reached 134 BPM during the two carries ... which were performed at a very slow and labored pace”). Ladder climbing, balance related tasks, bending, repetitive and sustained squatting, and kneeling tasks all tallied poor performances despite appropriate effort from Madison. Id. at 203.
“Based upon subjective and objective data gathered during th[e] FCE,” Navarro concluded that Madison put forth “full and consistent effort” for each test. Doc, 26-2 at 192. He “was not only willing to attempt all tasks requested of him, but also willingly worked to his true physiologic abilities.” Id. “At no time did [Madison] display an exaggerated walking presentation, and his inabilities correlated with his case history and objective impairments discovered during the musculoskeletal assessment.” Id. at 202.
“[V]alidity tests” confirmed those observations. Id. at 192. “Accordingly, the physical demand level and the restricted tolerances documented in [the FCE],” concluded Navarro, “may be safely and reliably used for work recommendations.” Id. And because the range of motion and strength measurements during the FCE mirrored those from December 2014, it constituted an accurate measure of Madison’s functionality during that time period as well. See Barbu v. Life Ins. Co. of N. Am., 35 F.Supp.3d 274, 292-93 (E.D.N.Y. 2014) (FCE found reliable in part because its measurements were corroborated by “results of similar tests” performed earlier).
Put differently, the FCE, a reliable measure of the functional effects of Madison’s knee on his work abilities, illustrated that climbing into and out of a truck; lifting 70 pound hoses; crouching to the ground many times each day; and driving a stiff-clutched truck would all pose such substantial difficulties that Madison could not safely perform the material duties of his job. Doc. 26-3 at 181. Since he also continually received regular care from a physician for his knee injury during the relevant time period (December 14, 2014-September 9, 2015), Madison was eligible for long-term disability benefits. Id. GGL thus de novo erred in denying those benefits.
GGL insists that FCEs generally provide poor evidence of disability, and, more specifically, that Madison’s borders on irrelevance because it cannot provide disability evidence during the relevant time period (=December 2014, according to GGL). Doc. 25 at 20-21. Too, says GGL, a “claimant has control over how well he performs in” an FCE, thus reducing their diagnostic efficacy. Id. None of those contentions survives scrutiny.
First, FCEs do provide a reliable, objective -assessment of an individual’s work-related function level. See Wise v. Hartford Life & Accident Ins. Co., 360 F.Supp.2d 1310, 1326 (N.D. Ga. 2005) (FCEs “have , at times been described by courts in this.Circuit as “the best means of assessing an individual’s function level”); Brookbank v. Anthem Life Ins. Co., No. 1:15-cv-165, 2016 WL 1611380, at *13 (S.D. Ohio Apr. 20, 2016), adopted, 2016 WL 2853578 (S.D. Ohio May 16, 2016) (“In the ERISA context ... a ‘functional capacity evaluation, [is] a reliable and objective method of gauging the extent one can complete work-related tasks.’ ”). And physical therapists like Navarro are licensed, trained professionals well-suited to assessing whether a given injury impacts functional ability. Surgeons, by contrast, better evaluate whether a person has an injury and whether surgical intervention could aid recovery. For example, neurosurgeons may operate on brain aneurysms, but therapists (occupational and physical) work *1396with patients to recover their functionality after surgery.
GGL correctly notes that claimants can “sandbag” FCEs, but Madison’s evaluation contained safeguards against such shenanigans. His heart rate—something he couldn’t “fake”—during the administered tasks reflected “near-maximal to maximal effort during testing.” Doc. 26-2 at 197. A Ransford Body Drawing, McGill Pain Questionnaire, and a Dallas Pain Questionnaire—all tests designed to ferret out malingering—fell “within acceptable parameters.14 Id. at 204. Navarro’s conclusions— that “Madison’s presentation ... was overall genuine,” that the “FCE [was] valid,” and that “the physical demand level and restricted tolerances documented” could be used for work recommendations—therefore remain valid. Id.
Finally, GGL contends that knee decon-ditioning as a result of Madison’s failure to attend doctor-ordered physical therapy, not any underlying knee injury, explains the FCE results and voids his benefits eligibility under the Plan. Doc. 28 at 10-12. It’s true that Bakheet ordered therapy on December 24, 2014 (doc. 26-1 at 188), and that McBride advised the same on January 27, 2015. Doc. 26-2 at 7. Shin recommended a “quad strengthening regimen” on November 11, 2014 (doc. 26-1 at 162), while Chuachiaco included therapy as part of his treatment plan as early as January 30, 2015. Doc. 26-2 at 13.
But that’s not the full story. GGL omits that Madison did attend a full round of therapy after surgery and reported to Ku-mar on October 27, 2014 that it conferred limited benefits. See doc. 26-1 at 168 (“[Madison] has completed full PT and also received left knee few steroid shots with no benefit.”). It also forgets that McBride ordered Supartz injections at the same time she recommended therapy on January 27, 2015, and that Chuachiaco’s plan delayed therapy until after the injections had a chance to work. Doc. 26-2 at 13. The last shot came on March 3 (id. at 27), and, in accordance with Chuachiaco’s recommendation, Madison started therapy thirty days later. Id at 34. He only attended two sessions because “the knee pain got so excruciating ... [that] he had to stop the therapy.” Id. at 218. Thereafter, no physician ordered Madison to undergo additional PT.
On that record, the Court cannot conclude that Madison failed to follow his doctor’s orders, or that deconditioning necessarily underlay the FCE results. Perhaps it played some role, but as McBride, Shin, Chuachiaco, and Craven all found, Madison’s osteochondral defect also had something to say. Again, and most importantly, Madison’s FCE provides valid evidence (that GGL apparently never considered, see infra) that his injury rendered him “unable to do the Material and Substantial Duties of [his] Own Occupation.” Doc. 26-3 at 181.
GGL also leans heavily on Craven’s November 5, 2014 exam to argue that Madison’s injury fell short of creating disability. Of all his many doctor’s visits, only that exam showed no objective measurements of left knee deficiencies. See doc. 26-1 at 167 (symmetrical quad and range of motion measurements, and equal strength in *1397both knees). In view of the fifteen other visits that showed at least some deficiency, however, the November 5 visit must properly be seen as an outlier. Indeed, just six days later, Madison visited Shin and exhibited significantly worse range of motion (125 deg. flexion for Craven, 90 deg. for Shin), atrophy, and tenderness. Id. at 161. At a minimum, Craven’s November 5 results cannot overwhelm the contrary observations of multiple physicians, contrary diagnostic imaging, and an FCE that provides much more specific, relevant, and, once again, contrary data.
Regardless, Madison contends that Craven essentially reversed himself on January 29, 2015 after conducting a record examination at Chuachiaco’s request. Doc. 27-28. The Court agrees. Chuachiaco saw Madison on January 5, 2015 and that same day ordered “[rjeferral orthopedics, chart review.” Doc. 26-2 at 1. Craven conducted that review on January 29, 2015, noting that he “[a]gree[d] with most recent eval,” and that Madison did “not need outside evaluation” because he had had “3 ... that all agree.” Id at 2. Because Chuachiaco ordered it on January 5, however, Craven included his review note in that day’s progress notes. Id
The “most recent eval” prior to January 29, 2015, was McBride’s January 27, 2015 exam. And that concluded, after measuring reduced range of motion and strength, that Madison’s pain was “most likely [due] to symptomatic chronic osteochondral defect/chondromalacia of lateral femoral con-dyle and weakness.” Id. at 6-7. For Craven to agree with that, he necessarily had to abandon, at least as of January 27, 2015, his November 5, 2014 conclusion that Madison “has [n]o significant orthopedic findings to explain his diffuse knee pain.” Doc. 26-1 at 167. His November 5 exam, then, carries little to no weight when placed in its proper contextual balance.
Nor do the Kalen, Benson, or Andrews’ record reviews. None examined, much less treated, Madison. Their opinions thus carry less heft as a practical matter, though examining physicians do not “automatically” deserve “special weight.” Black & Decker Disability Plan v. Nord, 538 U.S. 822, 834, 123 S.Ct. 1965, 155 L.Ed.2d 1034 (2003). Kalen didn’t even have Madison’s records to review—she only examined GGL’s “clinical highlights” (which are not part of the record) and referral form. Doc. 26-2 at 48. That’s hardly enough information to draw any conclusion from, and certainly does not support broad brush statements like “[Madison] should have been able to perform his occupation before 11/2014.” Id. at 49.
Benson isn’t much better. Although he reviewed actual medical records, he apparently never considered McBride, Chuachia-co, or Craven’s January 2015 notes. Doc. 26-2 at 72. GGL blithely asserts that it “has ... confirmed that [Benson] ... received [them] ... before [his] review,” and that “[i]t is irrelevant that some pages were not listed on a particular section of [his] report.” Doc. 28 at 21. But saying it doesn’t make it so. Benson neither mentions nor lists the omitted records. He lists other records that he does not mention, and still others are listed and mentioned. The only logical conclusion is that he did not consider the January progress notes at all.15
*1398What’s more, and despite having access to 2015 records, all of which documented restricted range of motion and strength, Benson inexplicably emphasized Craven’s November 5, 2014 evaluation. Doc. 26-2 at 70, In fact, his only mention of later records is to incorrectly note that no post-November 5, 2014 provider gave “a firm explanation for [Madison’s] continued self-directed complaints of pain.” Id. at 70. That’s not true (McBride and Chuachiaco both attributed Madison’s pain and limitations to his osteochondral defect, see, e.g., doc. 26-2 at 7 (“Pain most likely [due] to symptomatic chronic osteochondral defect .... ”), nor is it the only factual inaccuracy in Benson’s report. He also incorrectly states that Madison’s strength “has repeatedly been rated as 5/5” (id. at 69), when in fact Craven’s November notes are the only notes to record full left quad strength (some measurements do show full hamstring strength, but neither party has suggested that strength measurement deserves the focus Madison’s quads received). The lack of records and factual inaccuracies fatally undermine Benson’s conclusions.
Anderson, the only reviewer who had access to the FCE, fares no better. He, like Kalen and Benson, concluded that “no clinically relevant medical evidence ... supports Madison’s] self-reported complaints.” Doc. 26-2 at 76. That’s just not true. The FCE, for example, is (a) clinically relevant, and (b) supports Madison’s complaints. Yet, Andrews didn’t see fit to even mention it, much less discount it. Indeed, he based that erroneous conclusion on nothing but Craven’s November 5, 2014 evaluation. Id. Such a report cannot under-gird a benefits determination.16
At bottom, none of GGL’s justifications for its benefits denial withstand de novo scrutiny. The Court therefore must decide whether GGL had discretion to deny benefits and, if so, whether it abused that discretion in doing so.
B. GGL’s Discretion
Because the Court finds that GGL was de novo wrong, it must assess whether the Plan afforded GGL decisional discretion. See Blankenship v. Metro. Life Ins. Co., 644 F.3d 1350, 1355 (11th Cir. 2011). It does. Under the heading “Discretionary Authority for Benefit Determination,” the Plan made clear that GGL “will make the final decision on claims for benefits.” Doc. 26-3 at 210. It had the “discretionary authority to interpret the terms and provisions of the Policy,” though that authority did not “limit the legal action that may be taken by an insured or beneficiary.” Id. The Court thus proceeds to step three of the Blankenship analysis.17
*1399C. Abuse of Discretion (Arbitrary and Capricious) Review
“Under the arbitrary and capricious standard of review, the plan administrator’s decision to deny benefits must be upheld so long as there is a ‘reasonable basis’ for the decision.” Oliver v. Coca Cola Co., 497 F.3d 1181, 1195 (11th Cir. 2007), reh’g granted, opinion vacated in part on other grounds, 506 F.3d 1316 (citing Jett v. Blue Cross & Blue Shield of Ala., Inc., 890 F.2d 1137, 1140 (11th Cir. 1989)). If a decision is “the product of a deliberate, principled reasoning process and supported by substantial evidence,” it must be upheld. Reid v. Metro. Life Ins. Co., 944 F.Supp.2d 1279, 1316 (N.D. Ga. 2013) (citing Metropolitan Life Ins. Co. v. Glenn, 554 U.S. 105, 128 S.Ct. 2343, 171 L.Ed.2d 299 (2008)). Substantial evidence in turn “is such evidence that a reasonable mind might accept as adequate to support the conclusion reached by the decision-maker and requires more than a scintilla but less than a preponderance.” Id.
GGL insists that “overwhelming evidence,” much less substantial, proves that “Madison no longer met the Plan’s definition of disability.” Doe. 28 at 14. That alone requires denying Madison relief. Id. Too, says GGL, it engaged in the required reasoned decisionmaking process by “repeatedly requesting] and receiving] updated medical evidence from Madison’s doctors.” Id. at 15. Were that not enough, it used “three independent reviewers ... which is a strong indicator of a comprehensive investigation.” Id
GGL could have used one hundred reviewers, but if they all failed to consider all record evidence, they engaged in unreasoned decisionmaking that, says Madison, cannot undergird a benefits decision. Doc. 21-1 at 35. In particular, GGL’s final decision denying his second appeal relied entirely on Andrews’ opinion which itself never addressed the FCE, the most pertinent medical record available. Id. at 35-36.
GGL, moreover, “ignored the medical opinion letter from Dr. Chuachiaco and the” January 2015 medical records. Doc. 21-1 at 37. Although administrator’s may disagree with such records, they cannot, according to Madison, ignore them entirely. Id They must address the evidence and GGL’s failure to do so, he says, resulted in “impermissible ... cherry picking.” Id. at 38. Even under arbitrary and capricious review, then, GGL’s denial must be reversed. Id. at 39.
Once again, the Court agrees. As noted above, the relevant final decision under review is GGL’s courtesy review denial. For the same reasons that decision and not its two earlier denials receives scrutiny, the FCE and other post-December 14, 2014 medical records are properly before the Court.
GGL’s internal deliberations and final decision on his courtesy review focused solely on two things: Andrews’ opinion and Craven’s November 5, 2014 notes. See doc. 26-3 at 42-43. They mentioned the Kumar, Shin, and December 24, 2014 Bak-heet visits, one Chuachiaco visit (March 12, 2015), and McBride’s Supartz injections, but only in the context of regurgitating *1400what Andrews listed (but did not analyze). Id. GGL never actually addressed those records, nor did it even mention many others, including the FCE. Andrews, on whose report GGL relies, didn’t either.
Instead, Andrews cited only Craven’s November 5, 2014 notes as support for his conclusion that “[t]here is no clinically relevant medical evidence to support [Madison’s] self-reported complaints.” Doc. 26-2 at 76. But that ignores the FCE’s existence, not to mention Craven’s contradictory January 29, 2015 note, neither of which Andrews acknowledged. GGL then relied exclusively on those same two fatally flawed reports in its own decision. Doc. 26-3 at 43.18 “Such a selective review of the evidence and reliance on a cold record file review by a non-examining doctor,” to the exclusion of plainly relevant and reliable clinical evidence like Madison’s FCE, “establishes that [GGL’s] decision was not made ‘rationally and in good faith’ and is therefore unreasonable.” Kinser v. Plans Admin. Comm. of Citigroup, Inc., 488 F.Supp.2d 1369, 1383 (M.D. Ga. 2007) (quoting Griffis v. Delta Family-Care Disability Plan, 723 F.2d 822, 825 (11th Cir. 1984)).19
IV. CONCLUSION
Because GGL abused its discretion in denying Madison’s long-term disability benefits claim, its motion for judgment on the pleadings is DENIED. Doc. 25. In light of GGL’s de novo error in denying benefits, Madison’s cross-motion accordingly is GRANTED. Doe. 21.
The parties joint motion to seal the administrative record (doc. 22) likewise is GRANTED. Their joint clarification motion is DENIED AS MOOT. Doc. 27. Madison shall submit a motion for attorney’s fees and benefits award, with appropriate supporting documentation, within 21 days of the date this Order is served.
SO ORDERED, this 21st day of December, 2016.
. All citations are to the electronic docket, and all page numbers are those imprinted by the Court’s docketing software.
. "Osteochondral defect” refers to an area of damage, either from a traumatic injury or underlying bone disorder, that involves both the cartilage that lines the end of abone, and the bone itself. See https ://stanfordhealthcare. org/medical-conditions/bones-joints-and-muscles/chon dral-osteochondral-defect.html.
."Flexion” refers to the action of bending the knee, while extension refers to straightening. Full flexion occurs when a person’s heel touches the back of their thigh, and full extension occurs when the knee joint is completely straight.
. Performed August 24, 2014, the MRI revealed a “full-thickness cartilage defect of the trochlea ... with mild underlying bone marrow edema." Doc. 26-1 at 173. It also identified a "shallow defect ... at the site of cartilage loss in the trochlea," "mild joint effusion,” “an incomplete radial tear of the [meniscus],” and the same "osteochondral defect” as identified in previous scans. Id at 173-74.
. The January 13, 2015 MRI showed the same "[c]hronic osteochondral defect” as the previous two scans. It also showed a sprain of the anterior cruciate ligament, but overall presented "[n]o significant change compared to 8/24/14.” Doc. 26-1 at 204-05.
. See doc. 26-2 at 192 ("Based upon subjective and objective data gathered during this FCE, as well as upon of the medical records provided, it was determined the client did perform a full and consistent effort today. Mr, Madison was not only willing to attempt all tasks requested of him, but also willingly worked to his true physiologic abilities in each test performed during the material and nonmaterial handling portions of today’s examination. Furthermore, validity tests administered were within acceptable parameters in all trials and the client demonstrated genuine organic impairments that are limiting his function.”)
. Sharon Blair’s quest for long-term disability benefits spawned two published district court decisions and an unpublished Eleventh Circuit opinion. The procedural ins and outs of those decisions (Blair filed two lawsuits, one before and one after the appellate decision, based on the same set of facts), however, have *1391no bearing on the legal principles they stand for and how they apply to this case.
.GGL informed Madison that he had “the right to bring action in federal court under ERISA Section 502(a) (1) (B)“ when it denied his first appeal, but it never expressly stated that he had exhausted his administrative review rights (though he in fact had). Doc. 26-2 at 130.
. To satisfy the “full and fair review” requirement, ERISA administrators need only afford one level of appellate review. See Blair, 955 F.Supp.2d at 1239 (citing 29 C.F.R. § 2560.503-l(h)).
. Though no court appears to have directly addressed this issue, other decisions have at least implicitly followed this Court’s reasoning. See, e.g., W. v. Empire Healthchoice As*1392surance, Inc., No. 15-CIV-5250 (CM), 2016 WL 5115496, at *17 (S.D.N.Y. Sept. 15, 2016) (administrator's voluntary second review, which included records compiled after its previous final decision, "withstands arbitrary and capricious review”).
. disputes. This information comes from Madison’s declaration, which GGL never
. To arrive at seven, GGL counted the three record reviewers it employed (Kalen, Benson, and Andrews); Craven’s November 5, 2014 visit; Shin’s November 11, 2015 visit; and Kumar’s neurological exam on October 27, 2014. Doc. 25 at 16-17.
. He also saw Kumar, a neurologist, on October 27, 2014. Doc. 26-1 at 171. But that visit occurred only to rule out neurological bases for Madison’s symptoms, not to evaluate his underlying orthopaedic condition as did every other doctor visit. Kumar’s findings thus have little relevance in determining whether Madison suffered an orthopaedic injury that prevented him from performing his job’s duties. Doc. 26-3 at 641. Still, Kumar did note that Madison’s “clinical symptoms are of left knee origin,” lending some weight to the conclusion that an orthopaedic injury underlay his pain reports.
. GGL implies that Madison could have gamed each of these validity measures. Doc. 28 at 11 n.4. Yes, they are written tests and are thus by their very nature susceptible to respondent interference. But such tests contain built in safeguards to weed out answers and patterns of answers that aim to subvert truth. GGL presents nothing but innuendo to suggest those mechanisms failed, and Navarro, the FCE administrator, concluded they did not. The Court sees no reason to side with lawyer argument over a medical expert’s reasoned opinion.
. Madison makes much ado about GGL’s initial failure to include those January visit notes in the administrative record. Their omission (they have since been added) from the record and from Benson’s list, insists Madison, points to nefarious purposes. GGL insists that an administrative oversight explains their initial absence.
Oversight or ham-handed attempt at deception, the records are before the Court now *1398and their initial omission caused the Court no grief. And because Madison ultimately prevails, he suffered no prejudice either (query whether prejudice would have attached had he lost, since it appears that Madison’s attorney possessed the records as early as August 2015, well before he filed suit). See doc. 26-2 at 183-85).
. This Court is not the first to reject a William Andrews opinion because he inappropriately addressed an FCE. In Caesar v. Hartford Life and Acc. Ins. Co., Hartford primarily relied on an Andrews report that rejected an FCE to deny a benefits claim. 464 Fed.Appx. 431, 434 (6th Cir. 2012). The report, noted the court, offered only a "factually incorrect statement” in support of its FCE opinion. Id. at 435. That "rejection of the FCE without a reasoned explanation” sub-marined Andrews’ opinion and, in turn, supported "a finding that the termination of Caesar’s LTD benefits was arbitrary and capricious.” Id So too here.
. Madison urges the Court to apply a de novo standard of review, but never offers any evidence or argumentation disputing GGL’s discretion assertion. Instead, he claims that GGL used an affiliate’s employee to make the benefits decision without expressly delegating *1399its discretion and thus that de novo review should apply. Doc. 29 at 18 (citing Anderson v. Unum Life Ins. Co. of Am., 414 F.Supp.2d 1079, 1098 (M.D. Ala. 2006). But only GGL sent Madison letters, not the affiliate. And only GGL employees communicated with Madison. At no point did the affiliate purport to make a benefits decision. Balanced against the Plan’s clear conferral of discretion, those facts push the Court to evaluate GGL’s decision under the abuse of discretion standard in Blankenship step 3. Blankenship, 644 F.3d at 1355.
. GGL’s entire, unedited analysis supporting its courtesy review denial:
Dr. Andrews advised the claimant did not have any restrictions and/or limitations. Dr. Craven previously released the claimant to return to work after he was noted to have full range of motion, normal motor function and no quadriceps atrophy. Dr. Andrews reported the claimant had reached maximum medical improvement. Dr. Andrews opined there was no clinically relevant medical evidence to support the claimant’s self-reported complaints or to justify any work restrictions and/or limitations that would have impeded his ability to perform his job from 12/16/14 through the present.
Doc. 26-3 at 43.
. Since the Court finds that GGL arbitrarily and capriciously denied benefits, it need not decide whether GGL labored under a conflict of interest when making its decision. See Blankenship, 644 F.3d at 1355.