Neither the Cortex or Axe FX provide you with a recording of the reference amp either, just like the Kemper, so the argument doesn't really make sense.
I never suggested or claimed a QC capture by itself is sufficient to demonstrate whether the unit is accurate. Like the Kemper, you can find multiple comparisons of QC captures vs. the reference amps, though from what I've heard, the QC isn't completely accurate either. When compared to Kemper profiles of some of the same source amps, it sounds a bit closer in a lot of cases, though.
As far as the Axe-Fx, there are at least some comparisons to the actual amps that a few of the models are based on, and they sound extremely close. While the Axe-Fx modeling algorithms may not be 100% accurate, the difference is Fractal Audio continues to refine them. In the case of Kemper, Christoph has stated that while the profiling process isn't quite perfect, there's virtually no room to improve it.
For me, the point of the discussion is that a KPA2 could potentially benefit from more accurate profiling.
There are real world users like Andy Sneap and Michael Wagener who say the Kemper is accurate. In that sense, it would take a very large leap of faith to accept that somehow they don't know what they are talking about.
We've already established that there are certain conditions in which the profiler has difficulty accurately capturing an amp. If you watch this interview with Sneap in 2012, he compares a Kemper profile he created with the actual amp. He claims he can't hear a difference, but there's an obvious discrepancy in the low-end around 100 Hz, and I'd bet any amount of money that a comparison of the frequency spectrum would demonstrate that. Christoph himself has stated that there can be a difference in the bass response sometimes.
The accuracy of the profile is contingent on the user's ability to mic up the amp properly, as well as adjust the preamp and other elements in the signal chain in a way that will result in a 1:1 profile.
This can account for some issues, I agree. However, the ability to accurately capture a profile isn't wholly dependent on the competence of the user, as evidenced by the comment from Christoph referenced above.
I am curious to know what amps you had that couldn't be profiled.
As mentioned previously, it really makes no difference which amp(s) any single user has had difficulty profiling, but rather the number of collective examples on the whole. More to the point, what's really relevant is that the profiling process has been established as having limitations and that it's not always accurate. It should be taken for granted by now that the profiler doesn't always yield a perfectly accurate capture, irrespective of the competence of the user.
Again, the point for me is that a KPA2 would benefit from improved (ie. more accurate) profiling, in my opinion.