Before I begin, I just want to point out that I have no idea how anything works, have zero audio engineering or programming experience, and am 99.99% ignorant of most things in the universe.
So from my understanding, the Kemper creates profiles by sending a series of test signals to the amp for it to interpret, then science happens, and then out pops a profile. As we all know, one of the limits to this process is that the Kemper doesn't learn how the reference amp's EQ controls work or interact other than how they're set while the amp is profiled. Same goes for gain.
But what if the Kemper could learn those things.
The idea I have in mind, is that the Kemper would send out a test signal targeting a specific frequency range, and then you would sweep the EQ knob for that frequency on the reference amp from it's minimum to maximum for the Kemper to interpret. For example, let's say you have an amp with your standard bass-mid-treb controls. You would tell the Kemper to target those frequencies individually, at which point it would send out the appropriate signal and prompt you to start sweeping the range of the EQ knob. This process would be repeated for each EQ control. Gain could be profiled in a similar fashion.
So basically, it would be amp modeling by way of profiling, if that makes any sense. Of course, if this doesn't make sense, remember that I know nothing!