Often I find that adjusting the gain in a profile of an amp doesn't behave in a similar manner to my real amp.
Correct me if I'm wrong CK, but I think that's because it's a sort of linear add/remove "distortion characteristic" knob based on the profile, so its' either multiplying the distortion convolution strength, or it's effectively trying to just reduce/add volume on the input signal going in to the amp sim (thus using the profiles own measured response), a bit like doing the same on the guitar pot.
So my thought on a solution is : what if you offered the ability to make another profile (or maybe even several profiles) of the same amp, but with different gain amounts.
Then the gain dial behavior could be calculated based on that. Raising gain would then blend/morph the distortion characteristic of the amp towards the higher gain profile, and allow it to extrapolate gain beyond that, and reducing would move the model towards the lower gain profile. This way the gain would always be super accurate, and could potentially even be used as a method to refine profiles to be more accurate for dynamic playing.
It could be a part of the "refining" process, just a matter of adjusting the gain dial on the Kemper to automatically let it know that you're about to refine with a different gain level on the amp and take note only of the values relevant to that setting only (you could even extend the same workflow to profiling the whole tone stack, e.g. adjusting the low, mid or high and refining and the Kemper could use the difference to work out the way the tone stack sounds.