I think of OTPs as a form of time-shifting. Think of it this way: if you and your correspondent have viable OTPs, that implies that, at some point in the past, you securely communicated that OTP. Since it is (at least) the length of a future message, you could have just passed a secure message then. Instead, you passed something that enables passing a secure message now.
I don't see a situation being frequent where you'd know the message weeks or months in advance of when you send it. If we follow your scheme, there's no encryption; we just meet up when we need to communicate. What if we're on opposite sides of the earth? If we had exchanged keys months earlier, this wouldn't be a problem.
The fact that your key material is the same size as the message means all your security is in the key, thus in your key distribution method. Among other problems, this means no rekeying without replicating your original hand-off - you can communicate exactly as many bits as your previously securely exchanged, no more.
...erm, isn't that the point?
Yes, with OTP you need to exchange keys and this shifts all of the security to key exchange, and while this wouldn't be a problem in other ciphers, it's a problem in OTP because of its other properties and shortfalls.
So at most I feel as though it's a security/convenience tradeoff; a tradeoff that's substantial or even dangerous in certain situations.
Simplistic interpretations are the bane of security engineering, especially where human factors are involved.