Slight Problem/Question about Bayesian filters

[fluffy <fl...@thesticks.com>]

 Hello,

 I have been running Spamassassin for quiet sometime now.. And very 
 happy with it and what it is doing. However I have lately been trying to 
 lock down a little bit more and get they bayesian filters going.. I have 
 saved all my spam mail in a SPAM folder and been running

 sa-learn --ham --no-sync --dir /home/username/Maildir/.HAM/cur/
 sa-learn --spam --no-sync --dir /home/username/Maildir/.SPAM/cur/
 sa-learn --sync

 And it says it learned HAM and SPAM and how many.. And I thought all 
 was working fine.. However when I put a message in there from lets say 
 SPAMMER01. This email is identical each time. I add it to my SPAM 
 directory and run sa-learn.

 I would expect if I where to do a check on it doing

 spamassassin -D < mailmessage

 I would tell me SPAM this time.. Yet it does not. It still rates it as 
 a 1.79

 I will also add that I read and it is said "It depends on what user you 
 do the sa-learn is what it will filter to". So I also run sa-learn as 
 that user as well instead of root and still the same thing.

 Then I read that it needs to be done as the same user as spamassassin 
 is running. So I run it as user spamd now as well. With still no luck.

 So now that I have said all that I have a couple questions.

 #1 Can I make these "learning" GLOBAL for all email boxes comming into 
 my system?
 #2 How do I make it so when I say HERE IS A SPAM message make it spam?

 Thanks for your time..