It's not in the SMTP spec because SMTP doesn't need to know about it. Your objection is correct but in fact that's the desireable behavior -- senders can't assume that foo+bar and foo+qux are the same person. If they could make that assumption, subaddressing would be defeated as a useful filter signal. As it is, it's not just plus signs -- qmail has been using '-' as the subadress separator for decades.
Subaddressing is standardized in RFC 5233 as a filtering signal.
As far as I know it’s not standardized at all; it’s not in the SMTP spec. It’s actually annoying that some people assume this is standard because they can break functionality, like assuming foo+bar@ is the same person as foo+qux@ when it’s not guaranteed to be.