IMHO is the reason for the great sound improvement KPA vs AxeFx not the underlying amp modeling but the interaction of all components.
The AxeFx concept must model the amp and cabinet separately - and then trying to simulate the interaction with parameters like speaker impedance, speaker distortion and many more.
The KPA can capture all this in context and adjust the remaining differences during the refine process. By this we come this close to the real thing.
When profiling amp and speaker in two steps - then we loose all this - and even worse, do not have any of the AxeFx parameters to compensate e.g. the wrong speaker resonance.
The result may sound nice for some - but the level of realism is not even on AxeFx level.
For a great KPA profile we need all amp, speaker, microphone, placement, room, preamp,... to be great - and that is the reason why a lot of the stock profiles sound this great.
As for my profiles:
I do the same and in addition do I create many profiles of each amp(many mics, settings, all channels...) - and need about one week for a set of them.
If this is worth the price of a set of guitar strings has everybody to decide for himself. It looks as if a lot KPA users love this service.