I've got an old Trimble GPS setup - a classic Zephyr Geodetic, Trimark III, 5700 base with a 5800 rover. I'm pretty much just using it for grading and building/utility location as we slowly build out our campus master plan. Lately I've been spraying grading boundaries and shape lines for a couple excavators we have working on a detention basin on campus.
I usually set up on an unknown point and calibrate the site to 2 known points (one mag nail in rock and one gin spike in pavement), using just one for vertical datum. I get what I think are good results. I measure HI (which I don't believe matters when set up on an unknown, unmarked point), I watch my rod height and each time out I finish with a check in to a tack and hub that I set several months ago. Yet, having been out there 5 times now I've literally gotten 5 different elevations at that hub, varying by up to 0.30'. Horizontal seems to be consistently accurate. I'm not using a GEOID or SPC, but just set no datum at 1.000 correction. Fortunately, for hammering a giant hole in solid rock, a few tenths here and there won't matter, but certainly will if I want to set curb form later.
Any ideas what I might be missing, how I could pinpoint the problem or possibilities of old instrument errors?
Mahalo, LSU, for all your knowledge and community service.