Obviously it is the ratio that matters (although I disagree that the data gives 0.25 for Pr (B ~ A)-- a more justifiable upper bound would be 0.35 - 0.4, given the limited amount of data available)
You state «a more justifiable upper bound would be 0.35 - 0.4, given the limited amount of data available)» but there's no justification for using an upper bound here.