Given two disagreeing polls, one little & imprecise but taken at facial area-value, and the other significant & specific but with a significant chance of currently being completely mistaken, what is the correct Bayesian product to update on these two datapoints?