I think that the filters you mention are an important mechanism here. These continuous variables like IQ or test scores get mapped onto binary ones (filters/gates) as you make it, or don't, into schools or jobs or other opportunities. It's not necessarily a linear or symmetrical mapping, either.

For example, when I was in graduate school, the chair was discussing how high GRE scores were not very strong indicators of 'success' in grad school (i.e. doing good research, publishing it, and getting out in a reasonable time). However, low GRE scores were much stronger indicators of 'failure' given these criteria.

IQ probably maps pretty linearly to GRE scores (for most of the small subset of the population who would ever take such a test), but with regards to scientific progress there are a series of thresholds that are more important (grad school, faculty hiring, securing funding, etc. etc.). These thresholds are really about opportunity and access.