There are two things that appear to be lacking at least from reading that page and perhaps from his whole analysis: a large sample size and a dynamic test.
First, if I just take all of it at face value, as it looks like he wants me to do, there were less than 10 rounds shot as part of this test. I think it should have been dozens or hundreds for each scenario. Now maybe he did that and chose not to call it out, but why wouldn't he add "Oh by the way we did this 100 times", since it would bolster his argument.
Second, like many catastrophic failures in general, I think this is a systems issue with several things stacked on top of each other until the sum of them all of them tips it over: ammo quality to be sure, gun cleanliness, barrel support, this gun's ability to activate the firing pin when not fully in battery, gun handling, gun mods folks have done, etc etc. I think that's what's going on in competition where we hear of so many of these failures happening.
While his isolation of one aspect is certainly valuable in obtaining data about this problem, it's not enough to tell the whole story and not enough to be accepted as conclusive.