They do survey people who rent as well as own, so they are measuring the market directly for that portion.
As far as the owners go, I think you're disregarding the fact that the specific houses that are not being rented do not have more authoritative rents. If you use actual rental data, then you have to have a model using assumptions to map one population to another, which seems fraught with potential problems.
The point of a survey is to collect survey data, and if they don't do it, then it is that much worse of a survey, because there is no 100% substitute.
Not 100%, but I bet doing something as simple as looking at the bank accounts of the selected groups in aggregate (over a long time period) and adding in things like tax returns could get quite a bit of the way there.
To be honest, I'm surprised something like mint.com or Personal Capital don't offer their own interest rate measures.