I finally got around to trying this out for myself. I rode a 25-meter course using Neville's method and a 200-meter course with normal riding. I used rim readings on all the rides, doing my best to estimate to the nearest thousandth of a revolution.
Here's the results