Well, I explain why the beam formula makes sense in that blog post (asymptotically going to the 25% cap) and the graph shows the line in comparison to real world combat log data points. So far I have not found compelling evidence that effects beyond the formula itself have an effect on the hitrate calculations, so I stick with my theory that the underlying formula itself is different.

It's entirely possible that the formula is wrong and it's just a coincidence that it approximates the data points so well, but it is a fact that (unbuffed) beam accuracy is higher than cannon accuracy. I cannot rule out that there is a different source for the difference, but I am absolutely certain that this effect exists and is not just a display error / problem with the parser as Borticus currently believes.

(To confirm the validity of the suspected beam formula, one would need to run several tests in a controlled environment for 2h+ to get reliable data points. This makes investigating this thing a bit difficult and also explains why the data points are not exactly on the calculated graph.)