I don't see why you feel the Protoge is the problem. From your description of the course, I would expect different counts each mile. It sounds like it is next to impossible to ride the exact route from each direction (assuming you did ride each direction). Even if you rode the same direction, I wouldn't expect exact readings, unless it was a straight course. And, I doubt that the 9 miles was straight - you are in Illinois, not Nevada.
You only had two separate miles that were off by more than a revolution, which, given the pavement you describe, isn't bad.
I use an electronic counter, and find it quite accurate. My only qualm with electronic counters is that if you stop with the magnet too close to the sensor, you will get spurious incrementing. But, once I discovered that, I don't have that issue.
I think your readings are well within reason, all things considered. And, I don't think a JO/JR/ReigelSuperDuperCourseClicker would give you any more accuracy in this situation.
Another possible factor is that Cal #1 was off by a small amount, giving you a very-slightly-longer mile than you should have had. It doesn't take much change in calibration to vary the mile by .15 revolution.