I've had zero deliverability issues.
edit: to be clear, the spam detection works well, and if you're going to make an error, it's far less annoying to bias (as they do) towards a few spam showing up in your inbox rather than actual messages getting lost in spam. Also, they show the spam score, and you can hand adjust your spam threshold if you want.
That's because it's not spam. I'm pretty sure there's a blog post coming soon about the difference between spam and "email I don't want to receive". If you just don't want to hear from a particular email address, the way to stop it is to put an explicit discard rule for that address in your Rules screen.
Likely changes for next year include making a much easier UI option for "don't show me email from this address again", because that's a fairly common request for situations like this where you have a vexatious sender. I suspect it will be implemented as an addressbook group - the anti-whitelist!
The student loan confirmation is sent to spam every month. I mark it as not spam.
In their defense, the message is full of all caps words that look like a sentence formatted from a sql query with little attention paid to formatting plus it has all sorts of spam words (bank, payment confirmation, loan, etc).
I would also be afraid one day if Google sees people fleeing to them, they'd come in and buy them. I'd almost whish there was a public trust or non-profit who would run an email server. Post office is a government service (for better or for worse), email is a bit like what regular mail was 100 years ago.