Your argument concerning the testimony of Deputy Fire Marshall Vasquez about his investigative record is based on what appears to be a misreading of the report's claim.  The authors first note that Vasquez must have overstated at trial the proportion of fires he's investigated that were determined to be arson, citing Texas statistics (Vasquez claimed at trial that, of the "1,200-1,500" fires he had investigated, "most all of them" were arson, despite the fact that his own office believed the incidences of arson to be about 50% in fires it investigated).  His subsequent statement about the injury rates in 'arsons' he's investigated--that about 50% of his 'arson' cases featured an injury or death--is almost impossibly high given the fact that, nationally, the rate of injury in fires (accidental and arson combined) is only around 1.45%.  Given that Mr. Vasquez is either exaggerating about the proportion of fires he's investigated that turned out to be arson or else he's simply the unluckiest investigator in the nation to have caught so many arson cases (Occam's Razor says the former is probably correct), and given that fire burns the same in Texas as it does in the rest of the country, it's perfectly reasonable to compare his claims in light of available statistical evidence and perform a basic "smell test."  (Mr. Vasquez's subsequent testimony that "all fire burns up" further calls his professional competence into question, since, even by the science of 1991, that was known to be an incorrect statement).

Generally, arsonists are out to create property damage, not injure or kill people (in recent decades, between 40-50% of arson arrests have been of juveniles under the age of 18), so the idea that the injury rate in legitimate arsons would be high enough to offset the difference from the injury rate in all fires, accidental and incendiary combined, to the degree that Mr. Vasquez cited is a bit of a tough sell (as is the idea that the rate of injuries in arson fires alone would come close to approaching the statistics cited by Mr. Vasquez--the leading cause of fires that result in civilian deaths is smoking, and civilian injuries and deaths happen almost 2/3rds of the time in homes with no functioning smoke alarms, according to the NFPA.  And yes, those are national statistics, which I have every reason to expect would apply in Texas).

Regarding the ceiling in the front bedroom, the report discusses the progression of the fire regarding the ceiling and floor in post-flashover conditions.  Were the room to have been post-flashover, however, Willingham would never have been able to get in the door because of the radiant heat: if his recollection is accurate, the room must have been PRE-flashover, meaning neither the floor nor the ceiling would have been burning.  Therefore, if he was indeed struck by falling debris, it's likely to have been from something other than the ceiling.  It's also possible that he misinterpreted the heat of the fire (which was already high enough to ignite his hair near the ceiling) as something falling onto him.  Putting it another way: his description could match what someone would feel if they were at the threshold of a room as it became fully involved.  What his testimony could not match, however, would be what would have happened were an accelerant used on the floor, which would have created a burning fire at his feet rather than an "orange glow" on the ceiling as he described (which would fit with what a pre-flashover room would look like).  And if ANYONE, defendant or state's witness, made manifestly false statements about one thing, then their other statements should of course be questioned, which is why examining the science is so important in a case where eyewitness and expert testimony provide conflicting accounts--in this case, all questionable testimony aside, the science doesn't indicate arson.

There is no dispute that a chemical that could have been used as an accelerant was present.  The issue is that the accelerant was claimed by the state to have been used in a way that is simply impossible to prove (and, in some ways, impossible to have happened), and thus the significance of the accelerant's presence is lessened by the fact that it doesn't fit with the theory put forth by the state, so the presumption of innocence should stand.  The accelerant could not have burned under a threshold or under floor tiles, which was the state's way of explaining an irregular burn pattern.  The pattern that was attributed to the accelerant was likely a result of flashover.  The presence of an accelerant is not itself damning in a fire investigation (one thing my father pioneered was determining a technique for differentiating between the residue of burned asphalt found in roofing materials and heavy petroleum distillates like kerosene, which were often mistaken in gas chromatographs up through the 1990s), so the fact that its presence only fit with either an unscientific theory proffered by the state, or an innocent explanation that fit the science and the events on the scene, actually weakens the state's case, rather than strengthen it.

The human testimony was conflicting--one witness, Burvin Smith, reported that he had to restrain a hysterical Willingham from going onto his front porch (where the only accessible door back inside was).  Nobody denies that Willingham's behavior was erratic, but "erratic" is not the same as "damning."  (For example, another case where junk science was used in a prosecution featured an accused woman who, after calling the fire department to respond to the fire that was killing her husband in their work shed, proceeded to lock her house up behind her and wait in the yard.)  Moving a car away from a fire could show either presence of mind or unnatural concern under the circumstances, but neither is proof of innocence or guilt.

(As to the claim of his ex-wife that he confessed to her: what exactly would the motive to have continued the charade have been after he had been strapped to the execution gurney?  He maintained his innocence until the moment of his death.  Further, like you noted above, conflicting accounts need to be given reduced weight, and the ex-wife's accounts conflicted in this case.)

There is no "lie" about Mr. Vasquez's personal experience in the report: there is, however, a very convincing statistical argument that Mr. Vasquez was exaggerating on the stand about his professional experience.  The very first thing taught in any statistics class is that, if the stats don't pass the smell test, they're probably wrong.  There is more than ample reason to doubt Mr. Vasquez's claims about his record, and pointing out how his statistics offered at trial vary wildly from state and national statistics is not "lying" by any stretch.  

Testifying "in good faith at the time of trial" does not mean the witness was not negligent, nor is it the same thing as "executing a man in good faith at the time of sentence."  Scientific rigor before the trial would have resulted in the fire being deemed accidental (as all fires in Texas must be considered unless evidence proves otherwise).  And professional ethics would dictate that the state fire marshall's office should have notified the court that the evidence their witness provided was unreliable in light of scientific advances generally accepted prior to the execution.  This is, in fact, precisely "a case of the government wantonly executing persons they know to be innocent," assuming "innocent until proven guilty" is still the presumption.

And before you continue implying that the expert review was biased, we should note that all four fire investigators in the report have served on the NFPA Technical Committee on Fire Investigations, which is responsible for setting the scientific standards that govern fire debris analysis, and indicates all are highly respected in their field.  Further, three of the investigators have law enforcement backgrounds (my father began at the Georgia Bureau of Investigation, Mr. Churchward was a police officer and fire investigator in Ft. Wayne, and Mr. Smith was a detective with the Arson and Bomb unit of the Tuscon P.D.), and have experience working fire investigations from both sides of the table.  None of the report's authors were compensated, and all signed sworn affidavits attesting to the quality of their work.