Are we dealing with semantics here? The way you phrased that, along with all the eye-rolling and "bizarre"'s, gives me the impression that we're somehow not talking about the same things. Your objections don't seem to correspond with the basic probability and risk measurements that I've been explaining to you. I'm not even sure what you mean by "overall average" and "applies to the subpopulation", since in the sense that risk is modified by certain characteristics that is correct (but obvious), but in the sense that you seem to mean it is incorrect.

Think of it as a simple conditional probability. Each of the conditionals (relating to "average risk" in your terminology) are operative on the populations. What you are calling "subpopulations" would be determined by the conditions.