First, let me agree with your latter point: we should more carefully document the locations where there is an almost continuous stream of pictures, and make a list of all of the runners to either side of Rossi that were seen. (Like, all of the 20 people before and after him in the finish results, for example.) The later in the course this is, the more convincing - and we should try to nail down the positions of these pictures along the course.
Second, I disagree with your criticism of the statistical argument about Mike only having one photo. You make a good point, that multiple photos of a person might be lost when a photographer takes a break, but *this is why we counted how many locations somebody was captured at, not how photographs of them were taken*. That is the key. So we have a list of how many locations each of the 100 runners around Rossi were captured at, and we can do statistics on that list. It follows a normal distribution, justifying a z-test, and this test implies that the probability of Mr. Rossi missing all locations by chance, assuming he actually ran the race, is about 1 in 11 thousand.
Sure, it's possible. But it didn't happen to any of those other 100 people, or even come close to happening to any of them. And if you crunch the numbers, the result matches the obvious intuition - that it's unlikely to be by chance.