"Anyone who knows my phone number can authenticate as me, and MITM is trivial."
Erm, nope. Have you read the RFC, or the standards around it? The purpose of the KMS etc is to stop that.
The same could be said for "Anyone who knows my website address can authenticate as me, and MITM is trivial." which is obviously untrue. _But_ there's a huge dependency on a trusted authority - a root CA for SSL, a KMS for MIKEY-SAKKE.
Don't get me wrong, MIKEY-SAKKE has some problems, and I'm wholly unconvinced by it.
Also, your argument that phone number != person's ID is wholly spurious. Yes, it's true, however it's irrelevant. If I phone your phone number, I expect to get you. If we know each other, we may authenticate with a voice challenge/response we've previously established. If we don't know each other, we'd need a key distribution mechanism.
All of which is irrelevant. The purpose of secure voice standards is to provide guarantees of the integrity and confidentiality of voice comms, and to mutually authenticate endpoints against a MITM attack. MIKEY-SAKKE provides this (for some value of 'provides'), as do other key agreement protocols such as ZRTP (instead of MIKEY)(which uses the user reading out a string).