The point of that experiment was that they had a chamber in a nearby university that needed to put something into their annual report to the funding agencies.
If the computers are not identical, it follows that they should give different results. It further follows that you can come up with the set of inputs that will maximize those differences for a give pair of computers. You can also come up with inputs to minimize them. The fact that you can game them to as much as an hour difference is mildly amusing, but that's about it.
So, you are saying that the results are completely meaningless and provide no useful information to consumers?
Do you think the dive profiles used in the testing were unrealistic?
Are you suggesting that they could have "gamed" the profiles a different way and the Most Liberal Computer and the Most Conservative Computer would have swapped positions?
Are you suggesting that if you took the most and least liberal computers tested and you went out and dived 4 dives a day for a month, and compared the NDLs you got along the way, there would be a random distribution of which computer gave more NDL on any given dive?
I'm trying to understand what makes the results completely useless for any purpose.
Maybe I just don't understand how computers work very well. I've only been developing software professionally for 32 years, so that is entirely possible. I could be entirely wrong. But, from what I have read, including but not limited to the reports I linked, I concluded that if I dived with the most and least liberal computers from that test, I would NOT see a random distribution of which computer gave me more NDL time. I concluded from that report that, statistically, I would see longer NDLs on the most liberal computer MOST of the time. How much longer would, of course, vary. But, I concluded that, in the right circumstances, it could be considerably longer. And even in the worst circumstances, it would almost never actually be less NDL by any significant (to me, of course - say 1 or 2 minutes) amount. I.e. I don't believe there would ever be any way to "game" the dive profiles to have the "most liberal" computer come up with 50 minutes less NDL than the "most conservative" computer.
Thus, I concluded that getting a more liberal computer would be pretty much a win-win. Sometimes, little or no benefit. Sometimes, big benefit. But almost never a drawback. In terms of allowed bottom time, of course.
How have I misunderstood the data or drawn inappropriate conclusions?