I suspect a large amount of the justification for Parry as the go-to secondary stat is based heavily on a small study I myself did back in late closed beta, which showed that Parry drastically improved parry and block rates on low level enemies. I never particularly meant the study to be anything more than a preliminary look at the Parry stat in the close-to-release game, but it found it's way onto Dancing Mad which was pretty much taken as gospel early in 2.0. The study had some pretty serious issues to be honest, as it was not preformed on high level monsters, nor did it provide a comprehensive look at a continuous spectrum of parry values. As such, I would not be at all surprised if the results I found are not completely incorrect and Parry is an awful stat to focus on.

That being said, your n is far too small. A comprehensive test of high level Parry would need tens of thousands of samples in order to be statistically meaningful. Also, your data would have to be normalized against a common control parry. You dismissed theoretical values and claimed that this experience shows a real world application of the numbers, but without the theory, your knowledge is isolated to this case. Did you parry less because you were fighting enemies of much higher level? Or was it because Parry uses a non-linear growth curve? Or is it because you had more Dex (which also modifies Parry) in one case than another?

Your findings have merit; I don't mean to say that they do not. Rather that it is far too early to draw a conclusion from them.