Andy,
All valid experiments should be repeatable, so I went ahead and threw something together to try to validate your numbers. It wasn't nearly as robust a setup as my previous measurements (fastened to a granite surface plate, no housing sections to compress, a weight on the other end of the cable, etc. I don't have any 9sp shifters other than the ones on my bikes, but the results still seem reasonable.
The setup:
Some observations:
1. I should have flipped the indicator around the other way, as you did.
2. Putting the jog on the indicator tip rather than the item attached to the cable prevents the cable from being being displaced as much by the force of the indicator. I haven't done the math to see how much this influences the results, but my guess is not much.
3. The 1-2-3 block secured to the water bottle mount on this frame is NOT a particularly rigid method of attachment, mostly because the bosses are not flat on this frame. It does however work for a quick setup, and I'm not planning to re-use it as you are.
4. I could introduce a full mm of movement by greatly increasing cable tension by hand. All of it appears to be in the housing run from the shifter to the downtube stops. I was a bit surprised simply because I have no shifting issues whatsoever on this bike.
Results next post.