What is different about Mollom’s spam filter?
Mollom uses a next generation, two-stage spam-filtering protocol. If Mollom's intelligent text analysis filter decides that some content is suspicious or if it is unsure whether the content is ham or spam, it asks the user to solve a CAPTCHA challenge before allowing the posting.
This challenge/response procedure will never block human users, and it allows us to set very tight spam filters and prevent almost all spam from getting through.
On a software level, there is also no "markHam" API call, making the service much more robust. A markHam call could allow spammers to trick users into marking spam content as ham. See this "The science behind Mollom: Spam vs. Ham" blog post for more information.
It may also be helpful to read through the tutorial for the Mollom Drupal client.