Having used ACT extensively with my static groups in the past, I'm skeptical of that actually happening to a degree that it invalidates the general value of the information and would need to see it demonstrated. Any variations should generally be small enough (often the result of DoT simulation variations, which I mentioned in my post, or other things, like ping or when the parser registers an individual user as in combat and thus starts recording, and so on) to still provide for a reliable benchmark.
Not that I take parser results as gospel--they have a lot of idiosyncrasies that anyone using them as a tool to gauge performance should keep in mind. But they're also the only tool we currently have to really gauge performance for individual DPS. Luckily, that's going to be changing soon, with the addition of the Training Hall (or whatever it's named).
