Technically SMTP can be decentralized in the way you describe. Anyone can run a MTA. In principle a MTA should try routing messages through alternate routes to the destination if the first route does not work; also messages are stored on the way if some routes are blocked.
The fact people do not currently use SMTP in a fully distributed fashion does not detract from the fact that e-mail can be (re-)made fully distributed if/when need be.
Everyone would have his own single point of failure, but it's still centralized (as I explain there) because SMTP is designed to rely on centralized services (DNS, which you can't practically avoid) and assumes servers which are single points of failure (even if there are millions of them, you can still be cut off from yours).
A government can disrupt email communications even if everyone has its own server because SMTP assumes single central points of delivery by design.
> if the first route does not work
Bitmessage does it even if the first route works. Which is exactly what makes email centralized, and Bitmessage not.
Imagine a malicious MTA saying "ok, got the message" but not relaying it. The email is gone forever and you wont even notice. In Bitmessage the message will reach its destination one way or another... by design!
If you have to hack email to be more like Bitmessage for it to be decentralized... what's exactly your point? :P
The fact people do not currently use SMTP in a fully distributed fashion does not detract from the fact that e-mail can be (re-)made fully distributed if/when need be.