Alan, 5% seems pretty loose. I used to compute statistical significance levels and required 99% repeated at least 3 times.

Roger