efa-project.org

This message have -4.51 score and Spamassassin Autolearn was NO This message have -9.21 score and Spamassassin Autolearn was YES My config set :
bayes_auto_learn_threshold_nonspam -1.00
ayes_auto_learn_threshold_spam 7.00

Why with -4.51 is not considered for autolearning ham if i have -1.00 as limit ?

I have seen that very often, on both ends, sometimes SPAM isn't learned although it is above my threshold but I haven't managed to spot a pattern or figure out why this happens.

oh, I have one idea though:

not sure where I read this but I think I read that if an email is BAYES_00/BAYES_99 it won't be auto-learnt. Not sure about this though.

Could you go back and double check the BAYES status of those 2 emails? Maybe the first one was BAYES_00?