This blog post argues so, but I have my doubts since the author works at a graduate program whose S-rankings are much better than its R-rankings. I have a feeling that R-rankings do capture some things that S-rankings don't capture. Professors who are far ahead of their time, for example, might be recognized as such, but I would expect that their papers probably won't get very high citation counts for some time.
No, both rankings are basically nonsense. Even if you agree with the NRC's choice of a single "quality" model across all intellectual disciplines, the rankings are based on horrendously incomplete and incorrect data. This is especially true in computer science.
Also, the claim in the blog post is an obvious joke. The S-rankings are "better" because writer's home department's S-ranking was better than its R-ranking.
Any ranking that tries to distill a large bunch of heterogeneous statistics into a single number should be treated with a healthy dose of skepticism. There is no inherent reason why a particular set of statistics should be weighted with one set of weights instead of another—and the two different weights could lead to widely divergent results.
That said, if you see a large number of rankings, and they all tend to have fairly similar results, there can be some predictive power in the collective set of rankings. And I would say it's a "loose" scheme at best: the difference between #1 and #2 on such lists is probably not indicative of much; the difference between #10 and #30 or #50 is much more meaningful.