How big are the differences a few micron that could be the uncertainty of measurement or something greater.
Are all the alignment and measured feature hits the same as any variation will have an effect on results, all features must have identical setups to have good correlation.