a good question?

I would try to use a Calibration Course on a level street. The hills would probably balance out over the long haul.

Now if the course is totally downhill, the it may be best to use a downhill Cal Course.

If you looked at a calibration course that had a 2 meter rise and fall over its 300 meter distance, you would probably perceive it as being almost flat. That equates to a 5k with 50 meters of total climb, which most people would consider to be moderately hilly.
A 5k with a rise and fall of 100 meters would be considered very hilly, and yet a cal course with a rise and fall of 4 meters would be perceived as having only a slight rise in the middle.
My point is that you don't have to deviate much from flat in a calibration course to be equivalent to a "very hilly" race course. And that many "flat" calibration courses probably already have a rise and fall of a meter or two.