SPAM learning not working.

I am running a GPL host and the same spam keeps getting thru even though it gets forwarded to the spamtrap email. I logged into the user interface (aka: port 20000) and used the 'delete spam' button instead and this is what i got:

Reporting this message to Razor and other SpamAssassin spam-blocking databases .. Learned tokens from 0 message(s) (0 message(s) examined) .. done, and deleted message too.

This is making me believe that the learning facilities are not working.

Has anyone else noticed this problem?