Here is some technical stuff that you may take for what it is worth to you: It is not uncommon for "normal" scales that work by deforming a spring that drives a dial to have a non-linear error. By non-linear the error isn't the same throughout the entire range. If it is off 1% and it is linear, it would be off 1 lb when weighing something 100 lbs, and 3 lbs at 300 etc. Non-linear error might be 1 lb at 100 and 20 lbs at 300.
Nancy LC brings up the issue of "precision" versus "accuracy". For trending, as long as the error is consistent, valid trends can be measured. And CMCole is on the right track for a calibration method, as long as the error is linear, if it is not, you would have to place an object of a known weight close to what you weigh in order to tell. With many scales where the center of the load is placed will effect the reading, so that is another variable that is hard to control.

I would bet my paycheck that the balance beam scale at your doctors is accurate (as well as precise).