Transport is the wrong layer to do encryption (much as I'd love to see more adoption of IPSec) because encryption and authentication are inextricably linked. Suppose you can confirm that you're absolutely, veritably talking to 174.16.23.46, and your messages can't be read by anyone but them. What good does that actually do you?
TLS is, despite all the problematic cipher suites, fundamentally sound infrastructure - which is why it hasn't been replaced. The same is true of OpenPGP. We don't need new protocols, just better implementations - an equivalent of LibreSSL for OpenPGP.