FYREFLYE  Did you train it to differentiate good mail from spam as per the instructions or did you just install it and run it?

Since the app has a 30 day trial period before you're asked to pay, why did you buy it?  
(Version 2.7.7)

CHUCK BRAMAN  I trained it to differentiate good mail from spam as per the instructions on an almost daily basis from September through early December. I bought it based on the consistent, raving, 5-star reviews on the assumption that after it was trained for a sufficient period of time it would be what those reviewers said it was. Maybe it is for them; it wasn't for me.

In defense of the developers, they wrote several very detailed emails to me trying to troubleshoot the issues once they saw my review. In the end, it amounted to explaining why the software sometimes makes mistakes, and admonitions on the importance of training it. I don't think they understand why it is so inaccurate after all this time, and neither do I, but that doesn't change my experience or evaluation. Take it for whatever it's worth to you. I would suggest anyone interested in it try it during the 30 day evaluation period and make a firm decision at the end of that period before paying for it based on their initial experience; if it's not working for you by then, don't assume it will later, as I did.  
(Version 2.7.7)

MICHAEL TSAI  I would go one step further and say that if you aren't getting good results after a week or so you should contact technical support. In most cases we can suggest a few tweaks to the setup that will make a big difference. It doesn't help either of us for you to suffer in silence for months and then post a negative review.

SpamSieve has a slider in the preferences where you can set it to be more conservative, to reduce the chance of a good message going into the Spam mailbox. I think this would have helped a lot, since SpamSieve didn't think the good messages were super spammy, just a little over the borderline.

With software that learns and adapts rather than working from a fixed set of rules, it can be hard to see the exact reasons for what it's doing. Every classification depends on all the messages that it's been trained with. There were certainly some anomalies with the training: messages from the same sender trained as spam and as good, mistakes that SpamSieve made that it wasn't told about for a while, etc. It may be that SpamSieve was working from some incorrect information or from a set of messages that was not very representative. If adjusting the aggressiveness slider in the preferences doesn't help, I would suggest resetting SpamSieve's corpus and re-training it with the misclassified messages and others like them.

Of course, I would also understand if after several months you don't want to put any more time into this. If you aren't happy with the software, we don't want your money. I cannot refund credit card charges after 60 days, but if you e-mail me your postal address I can send you a check.  
(Version 2.7.7)

MICHAEL TSAI  If you're having a problem with SpamSieve, please contact spamsieve@c-command.com.

That said, SpamSieve 2.7.7 should be compatible with Mac OS X 10.4 through 10.6 (including the latest developer seeds). You can reinstall the plug-in by choosing "Install Apple Mail Plug-In" from the SpamSieve menu.  
(Version 2.7.7)