Nice post here about the manner in which good fictional dialogue radically misrepresents the actual words exchanged by interlocutors. Her example is the original publication of the Nixon tapes, which led many people to wrongly assume that the Nixon administration was populated by subliterates. Linguists earned their pay in the 70s explaining to a bemused populace that we don't actually talk very much like we write at all, especially the way we write about our own talking.
I think it's much too quick to just say that this is just one of the ways that fiction misrepresents the actual world. If I remember right, David Lewis was actually onto this. For him the set of possible worlds determined by a fictional text are not the ones consistent with the words on the text, which (for the reasons Emily gives as well as the odder ones elaborated by theorists of unnatural narrative) would be shockingly weird worlds. Rather, they are ones where the story is* correctly told as true. So then the important question comes down to what it is to tell a story as true.
Unfortunately most, if not all, of the work in unnatural narratology (which includes a lot of brilliant text analysis) makes a hash of this precisely because they assume a naive version what it would be to correctly tell a story as known fact (they also typically assume a distinction between the natural and unnatural no less problematic than an American undergraduate writing their final ethics paper against gay marriage). Rather, what needs to be explored is the extent to which correct communication of facts about the world constitutively involves misrepresentation. But since the nobody is thinking much about the actual truth in fictional texts any more they don't use fiction as test cases for how this works.
It's weird that people in unnatural narratology miss this, because Lovecraft's attempted representation of the unrepresentable is interesting philosophically precisely because he might be teaching us a lot about the actual world. Moreover, this was clearly his gloomy intent.
[*Is, or can be? I'm almost certain that the former is Lewis' view. It's problematic in either case because lots of stories are such that nobody in the diagetic realm is telling or could be telling the story in question.]