Nice analysis. Here's mine from last year: http://fixedandfloating.blogspot.com/2010/06/greatest-match-ever-some-a…
Sorry it's a lot of rambling at the start - I do get to the probability thing eventually.
As you say it's very sensitive to the value of p. I just used the value of p from that match (even though I didn't like doing so) because p depends a lot on who's on the other side of the net, and hey, there were a lot of points in just that match :)
I note that your probability is *given* that they get to a fifth set in the first place.
An interesting side note: they drew each other in the first round again this year. What were the odds of that?! Isner won again (a bit quicker this time).