“it must be that X” can be [used] to express high confidence in X. What’s striking me is that this is so even when, intuitively, my epistemic position isn’t strong enough to just assert X. Is that weird? ...
I know that sometimes, pragmatic rules prohibit me from asserting things that are entailed by other things that I can assert, like when I say “she’s more than four feet tall” when someone asks me how tall she is and I know that she’s 5′4″. But this seems worse: I’m prohibited by something like the maxim of quality from asserting something that is strictly stronger than something I am permitted to assert.
This strikes me as a special use of the modal 'must'. In saying, "it must be that X", we are not merely communicating the claim "it is the case that X" with additional modal force ("it is - indeed, must be - the case that X"). Rather, we seem to be saying something more like, "General considerations force me to conclude that X." We are indicating a certain indirectness of thought; a need for inference from the general to the specific. That is how we reached the conclusion ourselves, after all. By forcing the listener to undergo a similar process, we communicate that this is how we reached the conclusion ourselves. (We may need to add this "maxim of mirroring" to Grice's list!)
On this account, then, direct assertion ("X is the case") implicates possession of direct evidence. Note that introducing the 'must' - though semantically stronger - requires us to make an inferential step before we reach the simple conclusion that X is the case. By drawing out the listener's thoughts in this roundabout way, one mirrors the thoughts of the speaker. That is, you communicate that your grounds for believing X are indirect (and so perhaps insufficient for knowledge). In this way, it can be felicitous to make the strong claim "X must be the case" - even if you lack sufficient evidence for the weaker claim that X is actually true.
A problem remains. Suppose you don't know whether X is true, but may permissibly say 'it must be that X.' Assuming that knowledge is the norm of assertion, Jonathan points out that we must either (1) deny that 'it must be that X' entails X on the above use of 'must', or (2) claim that 'it must be that X' is not an assertion but "some more tentative thing." (I guess a third option would be to deny that knowledge is closed over entailment.) Which of these options most naturally fits the story I've told above?