It’s possible that many users classify them as junk and so the Bayesian filters learn from that. Plus they might actually be junk. The filters are usually content based.
Yep, my first reaction was, "seems to work as designed". The fix isn't to whitelist, the fix is to make their own emails more relevant (or easier to unsubscribe from).
I’ve known people who use the spam button for any email they do’t want right now. They say “it’s easier” than hitting the “unsbuscribe” link or they say “if I hit unsubscribe they’ll just send me more email”. It’s hard to maintain a rational filtering system under those circumstances.
I've seen faked brand emails hit my Gmail inbox with (presumably malicious) unsubscribe links, and I've definitely seen sites "forget" that I've unsubscribed.
It's a bit of a tragedy of the commons, but I think the state of email is such that simply reporting unwanted email as spam is a rational choice
