I have never been happy with having a great variation (more then 2-3 couts/mile)with the calibration. Over the past few years I have measured a new calibration course as close as possible to each race course. It does require a second person (or you can use Pete's single person method).
For most of my measurements in NYC, I measure in the middle of the night which often reduces the chance of variation. Try it sometimes - no vehicles, no runners, no bikers, - just you, the racoons, and yea - sometimes Jim G!