I've heard the Kemper does Marshall's very well. It could possibly lend to the theory that the KPA uses a template or source tone that may very well be based on a Marshall and refines from there, so the further you trend from that tone, the more pronounced differences become.
I think it depends on the Marshall. Some of the later JCM's maybe. I agree with the theory the KPA probably has templates it selects from and then applies EQ matching of sorts. If you know tube amp circuits then you are aware there are only a handful of different circuits out there that are all gain staged in different ways, combined with differing components (transformers, chokes, etc) and EQ frequencies, that all create different end results. Even circuit board layout impacts tone. I've read interviews where CK says he initially embarked down the path of modeling with intent to finally solve the digital modeler shortcomings but after a good while down that path he realized it would be too tedious to cover many amps. So my guess is he took that knowledge to create templates and then applied EQ matching of sorts (which under this theory is what the refining phase of profiling is likely based on). So yes, if this theory has any merit it is very possible the core templates are not always close enough to the amp to result in a proper duplicate profile. Then again, seeing as how the same issues (rasp/congestion) are surfacing across a variety of profiles (in varying degrees) I still lean more heavily toward there being something else going on in the core tone engine logic and/or hardware design.
Sonic