Certain arguments that I've used myself in the past have presupposed that back-reference to things one's interlocutor said standardly reference the interlocutor's expressed proposition. For instance, if George says "I am hungry" and you say "I believe that", you are not expressing a belief in your hunger, but George's—you are assenting to the proposition that he is hungry. I knew, of course, that there are cases where things aren't quite like that. For instance, you tell me on the phone "It's a cold day", I might say "Not here!" But of course the proposition you uttered includes a rigid reference to the place you uttered it (it's implicitly indexical), and it is just as true there as here. I took these exceptional cases to be "mere exceptions", where one references the sentence rather than the proposition, or cases to be handled by, say, positing a "relocation" operator that acts on propositions.
But I've been thinking about religious dialogue and coming across more cases, and ones that may be fairly important to our communication. Here is one example. My concept of marriage is such that it is analytic that marriage is not just a social status. But let's say you think that marriage just is a social status, and you know what my concept of marriage is. When I say "Fred and Jane got married last Saturday" then I am expressing a proposition that you believe to be false. For the proposition that I expressing is that they entered a marriage—i.e., entered into a marital relationship that is more than a social status. If standard back-reference is to the expressed proposition, then you should say "That's not so!" But instead you will say "Yes" or "That's true" or at most "Indeed, though I understand 'marriage' differently." Moreover, because differences in concept are quite common, this is a pretty normal situation.
It seems, then, that what I refer back to is your words, classified lexicographically (if you said, "Yesterday, Fred was at the bank", meaning a financial institution, it would be misleading for me to agree, when I believed that Fred spent all of yesterday by the side of a river), and with indexicals shifted. The effect of your "That's true" in these cases seems to be basically that of repeating (without plagiarism) my words, with indexical shifts, rather than asserting my proposition.
Perhaps, though, this is just another exceptional case, reflecting the fact that "That's true" in ordinary concepts means "That's close enough in the present context."
If, however, this is more than an exceptional case, then I may have been wrong in this paper of mine when I said we had a duty to use words in our interlocutor's sense.