This weekend I was able to move over to the next street beside mine to repeat the hill experiment on a street where the hill calibration course and flat calibration course had identical road surfaces. The courses were 180m long and the hilly course had about a 4m change in elevation. I conducted the experiment in the same way as the previous ones, with a total of 16 flat and 16 hill rides.
Again I get the average cal constant of uphill and downhill rides to be smaller than the average of the flat rides, although it is only 6 counts/km in this experiment compared to 10 in the previous two experiments. This change in magnitude may be due to the different surfaces on the hilly and flat courses in the previous tests.
Not sure why the flat out ride average was higher than the flat return ride average, but still, both flat ride averages were higher than both uphill and downhill averages.
After 3 separate experiments with a total of 96 rides, I have come to the conclusion that for ME riding MY bike, there is a hill effect. That effect is that averaging uphill and downhill rides on a hilly cal course will give me a lower cal constant that averaging rides on a flat cal course. This may not be true for others.
However, I do believe that for everyone it is a mistake to use a flat cal course when the race course is hilly. In this case, if there is a difference between flat and hilly cal courses, then it makes sense to use a hilly one. If there is no difference between flat and hilly cal courses, then using a hilly one won't hurt you.