False-positives and false-negatives with SpamAssassin

I just started using SpamAssassin and I'm experiencing a lot of False-Positives and False-Negatives. What can I do to make SpamAssassin more efficient?

You can radically improve SpamAssassin's results by ensuring you have blocklists (DNSBLs and URI DNSBLs) and networked plugins (e.g. Razor, Pyzor) and that you are actively training Bayes (which takes 200+ spams and 200+ hams).