It's more that you only need to collect 10% of the data to begin with, not necessarily throw away 90% of what you do have.
I could be wrong, though -- I only did this back-of-the-envelope a few months ago. I never really tried that hard to make it work.
This is most likely true. Additionally, raytracing just might produce samples that are too regular for this to work well. I haven't really thought about it :)