You're conflating two different estimates: Boisjoly was certain that the particular launch would fail, and a survey of engineers gave a 1-2% estimate for failure across all flights.

Both of those predictions turned out to be extraordinarily accurate.