Well, for now I'll stick by my subjective impression (which was fairly strong) that UpRev and the tuners are right: maps with a solid column (or more) of 3800 values at the end do get on the throttle quicker at lower RPMs when you stomp the pedal all the way.
The MPH and RPM data via Cipher simply isn't good enough to infer g's of accel at the fine scale needed to tell the difference between the stock and non-stock maps at low RPMs in the scenario above, and adding to the error is my inability to get the graph start points lined up precisely (same starting pedal pos + rpm, leading to same timestamp for first full-throttle sample).
I'll post up this chart here anyways so you can see what the data looks like, but the bottom line to me is "too noisy and inaccurate to tell anything from". MAP4 is stock, MAP3 is one of the 3800 maps (the 2300 one in this case, not that it matters which at full throttle):
