But it's still an approximation, not an estimate, right? (The fact that you can list the mean radius with sub-decimal precision suggests as much.)

To me, an estimate suggests that there's error bars; an approximation suggests that there's variance that we can quantify (or at least we're very confident about our error bars).