I'm confused - first you wanted education to subsidize research. Wouldn't that have been taking money away from education, back when you believed that was the status quo? It seems as if your fundamental assumption is that the status quo must be correct.
In any case, yes, I agree with you that most researchers will need to have some sort of undergrad education. I don't see any reason why that education should be provided by researchers - cheap adjunct professors seem to do a reasonably good job at it.
(I do agree that a PhD program will need to be taught by researchers, since it is basically just an apprenticeship.)
As for teaching, my experience was worth considerably less to my students than a better ability to explain simple concepts would have been. The fact that I developed a nice outgoing wave filter for the Schrodinger equation did not help my fashion business student who barely knew algebra. It didn't even help the top 2-3 students who were breezing through my class, who might (possibly) have understood the concept of a PDE, or at least nodded along as if they did.
What might have helped them was someone who (like them) struggled with calculus, mastered it, and knew where the difficult points were. It also might have helped to have someone who could get to know students, understand how they learn best, and then tailor the lessons to suit them. Or someone who can look at a sea of blank faces and determine when they are confused. Unfortunately, that person wasn't me. I know a number of people who are that person, but only a small fraction of them have published anything beyond their PhD thesis.