WHOA! I think I'm on to something. I just looked at data for several other bikes (zx10r, cbr500rr, harley street glide, honda shadow 750, etc.). Guess what? The shifts are all centered around peak HP! It seems more likely that this is not just coincidence. It might be a correct generalization to say that you should shift such the RPMs before/after the shift are centered around the peak HP RPMs.
After some more thinking, this could only be true if the HP curve is symmetrical in the vicinity of peak HP. All the bikes I looked at (including the FJR) had generally symmetrical HP curves in that area, which is why the before/after RPMs for each shift were generally centered around the peak HP RPM.
The BMW K1600GT, however, does not have a very symmetrical HP curve around its peak. The calculated optimal shift points also end up with RPM changes that are not centered around peak HP (and not even consistently centered around a single RPM value). I think this proves that there is no way to determine optimal shift points based purely on where peak TQ or HP occurs. The "centered around peak HP" technique is only a decent estimate, and is more accurate if the HP curve is more symmetrical near the peak.