Neil, if you are still willing, take as many 3790 as you have and record some tracks, so we could see what is typical deviation. Additionally you could swap nuvis in mounts. All this could help to establish some measurement reference.