A good email blacklist (such as the Spamhaus Zen list) will detect 90% of spam with no false positives and negligible load.
There is a lot of evidence that there are not many botnets and each one has little diversity of control. This observation does not generalize to other kinds of spam such as 419s.
1. What makes you claim blacklists are the "most-effective" form of filtering? My understanding is that other techniques like Bayesian filtering are.
Background: http://www.paulgraham.com/falsepositives.html
2. Saying that it only comes from a few sources is dangerous thinking. This type of thinking was used by politicians to get a ban on adult services from Craigslist ("get rid of it there and it goes away!").