I think you guys are thinking too much the "old school" way. There was a time when getting a line to hit a dot on a TV screen was thought to be impossible and not practical and needed to much data to do.
For me, the pinnacle of modeling is when an amp's total parameters are captured in full and digitized to a file, or if the modeler becomes sonically better than what any tube amp can ever do.
The only thing that can't be done is multiple profiles on the current architecture, Kemper have already stated this many times and no amount of A.I. will sort that.
So it either needs a new unit or as has been suggested applying the tone stack principles - which seems very logical. In fact seems similar to the Kones - applies a specific tone to emulate the speaker.
However, my point echoes the last bit in your quote - to my mind its already better than the amp its profiling because for example you can add more gain to say an AC30 than the original amp was ever capable of. Like you I only care about better sound rather than pure faithful reproduction of every parameter. The other point is, if there is a sweet spot where the tone controls interact nicely then profile that...its only if you want to change eq on the fly that it becomes "useful".
Like Alan, I don't want to guess what the eq block changes are goign to do BUT I do support an evolution like this purely because some people want to be able to fully emulate their amp or effects.