You are right, but these numbers are intrinsically affected by value-judgements - about how to integrate impacts over time, across different sectors, across rich and poor countries/communities and over probability of such impacts (risk aversion). It's not so much the science changing, but the values - hence political shifts. It would help if experts could separate these factors more clearly. For example people mention "the discount rate", but there is not just one - there is a (low) pure time preference for the whole world and higher rates for individuals and companies with finite lifetimes, also higher in rapidly developing countries (this does make sense, given a non-linear welfare function).