One more note, it looks like my first pass at determining cable pull per shift was repeatable within +/- 0.05 mm (0.002") across 5 runs. The instrument I was using has sub micron repeatability, so the variation was all my setup and the front shifter. The upshift data was noisier than the downshift data.
Something I have not seen in this thread is an idea of how close is close enough. +/- 0.1" is likely too coarse, and the 2 nm flatness of our interferometer calibration artifact is likely not needed.
The rear derailer adjustment screws I've measured have all been M5x0.8, and the smallest adjustment range is 1/8 turn, which equates to 0.1 mm as the smallest adjustment. A single detent of adjustment doesn't make much difference, so I'll argue anything under half of that is good enough for what we're doing. Parenthood has thrown that particular project on the back burner, so I probably won't have more data any time soon.