Can you provide some sample data? If you can do this for 3-5 messages, that may be enough to see a pattern. Here's what you need to do:
1) Find a message that was incorrectly evaluated in the Message List
2) Right-click on it and select:
AntiJunk -> Show Evaluation
3) Select and copy the text in the window and paste it. Be sure to censor any sensitive information -- email addresses in particular.
Putting the text into a code block helps format (ie. it's the 5th button above where you type your reply in the forum -- bold, italics, underline, quote, code, ...)
Here's an example of what you may expect:
Code: Select all
Score: -2.5
White List: score: -1.5
xx@email.informeddelivery.usps.com: (-1.5)
Rules: score: 0
USPS - retrieve entire: (-)
Bayesian: score: -1 0% (conf: 460 / 547 = 84.10%)
Explanation:
- The overall score was -2.5.
- The email address is in the whitelist, which uses the default score of -1.5
- I've created a custom rule to "retrieve entire" (I retrieve preview by default, but these usps messages always have attachments that I want to see); this rule does not affect the overall score; the weight is 0
- Bayesian calculated as legit with a score of -1
So the 3 things that hit: -1.5 + 0 + -1 = -2.5
I do wish that this window also displayed whether the -2.5 was legit/junk/unknown, but -- for now -- it does not.
P.S. Don't take a year to reply back; we'll get it figured out
