Spamassasin bayes behavior with spam messages

Hello everyone! I’m trying to set up spam filtering on my mail server. I enabled automatic learning in spamassasin using bayes. Everything is fine, except that even if bayes recognizes letters as spam by 99%, it adds only 5 points to the overall spam_score value. By default, spam_score should be 50% for a letter to end up in the spam folder, but all sorts of SPF, DKIm and other checks reduce this value. As a result, even if Bayes is 99% sure that it is spam, letters do not get more than 25-29% Spam_score. I would not like to change the default value, because I am afraid that the probability of incorrectly identifying ham letters will increase. I found another way, enabled the shortcurcuit plugin and the option in spamassasin that if Bayes score>99% it rejects the letter. But in this case I won’t see if the email was mistakenly recognized as spam. I would like to change this behavior and I see 2 possible ways:

  1. Change the added number of points in spamassasin for bayes (I didn’t find where to do this).
  2. Using the shortcurcuit plugin, do not discard spam emails, but move them directly to the spam folder.

I would like your help in understanding where I can change these settings. Thank you in advance.