I can't argue with what you are saying other than I have experienced the same until I changed varieties and then the calibration would fall apart.  Since we have a state approved scale for semi's and we weigh everyload we harvest we find these "absolute" errors quite often while still knowing that the yield monitor is finding relative spatial differences.  It is not all that frustrating until you want to analyze whole fields against other data sets and know the whole yield map is skewed to the variety that was in calibration vs the variety that was not.  That is why "easy" post calibration of yield data against real life legal tender data (scale weights) would be beneficial.  But there always seems to be more discussion about how good yield monitors are in the absolute rather than good in the relative.

 Also I think I said I have found good relative comparison values between "test plots" for say fertilizer treatments or seed treatments etc as long as they had the SAME variety.  BUT when I am looking at variety differences I will always use a scale that is calibrated to lbs rather than a yield monitor that derives it's calibration from density (test weights) as well as other moving targets in the system.

Arnie