I think the logic behind this dates back to the original measurement philosophy- "AVOID SHORT COURSES".
We use the longer of two measurements of a race course as one of our ways to insure that the course isn't short. That's also why we use the larger of the two calibration constants to calculate the course length.
Nobody sets records on calibration courses. There is no motivation for them to be anything but accurate, and the best way to do this is to average the measurments.