I occasionally receive commercial messages from legitimate senders using a new FROM name and address. These messages are often very spammy in character. For these situations, is it better not to train Spamsieve so as to avoid improperly biasing the corpus and, instead, add the sender to the whitelist and/or address book and remove the sender header terms from the corpus?
I see what the manual says (below) and recognize this probably isn’t a significant effect, but I’d like to establish a consistent practice that made sense in view of experience to-date.
p. 60 says “it is imperative that you correct SpamSieve when it makes a mistake; otherwise it will “learn” things that aren’t true and begin making predictions based on that incorrect information.” [Similar instructions appear elsewhere.]