I use the spamhaus DNSBL on my server. This rejects typically 3000 spams a week according to my weekly automated reports. Spamassassin in reject mode (scores > 10) gets rid of a further 200.
In filter mode, I use more agressive DNSBLs (including one I compile myself) and a lower Spamassassin threshhold score ( > 7.0). I get about 300 spams per week in my spam folder, where I check one line per email (sender and subject) twice a week for false positives, and I get about 1 FP per month there.
About 20 spam emails a week make it through to my inbox.
Many commercial email rejection/filtering services provide their customers with similar or slightly better performance here than I achieve myself.
To be able to improve upon this various incremental improvements in existing standards and software based approaches are possible including:
a. ISPs to implement standards such as automated activation of RFC2369 headers when someone clicks a "this is spam" button (based on subscriber regret). This would be better than AOL making these headers invisible and bouncing an anonymised and untraceable complaint to an abuse handler unable to remove the confirmed opt in with subscriber regret.
b. Better means of identifying IP addresses which should not be sending email directly across administrative or contractual boundaries in the first place, such that an ISP can mark all addresses other than their own mailservers unsuited by default with domain owners publishing CSV records ( http://www.bbiw.net/CSV/draft-ietf-marid-csv-dna-01.txt ).
c. DNS and email server software and services making implementation of standards such as those above and DomainKeys a lot easier.