I do not believe we are looking at true signal strength, but rather something more like an inverse error rate. So 100% should correspond to zero errors.
I do believe, as Sam said, that excess signal strength can overload the input stages of any tuner and cause the error rate to go up. But whether too low or too high, the "100%" reported by mopardude should go down. I think his 1 sec dropouts are caused by some other issue.