This is the catch. A confident prior that disagrees with the data drags the posterior away from the truth and can inflate the false-positive rate. It's exactly why the FDA asks sponsors to simulate operating characteristics and run sensitivity analyses before trusting the savings.