MailScanner 4.66.5 and SA 3.2.4 score/bayes issues

[Raymond H Thrush II]

Has anyone else noticed any changes in your SA scoring after upgrading 
to MS 4.66.5 and SA 3.2.4?

I have my scoring set to Bayes auto learn ham at -2.0 and Bayes auto 
learn spam at 30.0.  But since I have done the upgrade Tuesday I have 
been seeing significantly lower spam learns, while my average SA scores 
have dropped and the most hit score above my high spam is around 10.xxxx.

In MS I weight them <4 ok >4 is possible spam and >8 is delete( high spam )

My average score for high spam before the upgrade was around 23.xx and 
the average overall score was about 19.xx
now I am seeing about 19.xx for high avg score and 16.xx overall average.

It seems like all my rule sets are ok for SA using the newest sa-update 
set compiled,  when I grepped against the previous version 3.2.3 rules 
they only differ in the version tags.

Any clues?

I believe I need to tweak my auto learn spam settings based on what I am 
seeing.