When we measure a race course twice we use the shorter measurement to establish the length of the course not only to avoid short courses, but also because it is almost certainly the more accurate of the two measurements. There are many ways to end up with a long measurement (bad tangent riding, etc.) but it's not as easy to end up with an incorrect short measurement.
When we steel tape a calibration course there is no reason to believe that either the shorter or the longer measurement is more accurate than the other, so we average.
For calibration rides, the smallest count we get is also probably the most accurate, but we use the average because of the danger that using the smallest count calibration ride will result in a short course. In this case, we allow "avoid short courses" to override accuracy.
That's my take on things anyway.