A good way to assess errors in one's astrometry is to measure several images of a field, not just a single frame. Place the target of interest in a different location for each exposure (say halfway from the center toward each corner of the CCD for four exposures). Using USNO-A2.0, you'll get a slightly different reference net for each picture, and the different location of the stars with respect to the pixels will introduce very small differences in the reductions, plus the optical geometric errors particular to your telescope set-up (focus, tracking, seeing, tilt of the CCD, etc). By then lookig at the scatter in your several measurements, you'll have a much better estimate of what your real extenal errors are. Almost always the error estimate found from the residuals in the fit to the reference stars from a single image will underestimate the "true" external errors, typically by a factor of two or three. Thus claimed uncertainties in supernova positions (for example on recent IAU Circulars) of 0".2-0".3 are much too small, and are much larger, plainly shown when other observers measure the same object on different images, even when there are no problems with crowding, etc. \Brian