Fictions about Real Things

I am on the plane from New York back to Los Angeles (currently over Omaha), and since Google was good enough to sponsor free in-flight wifi for all Virgin America flights, I figured I'd take this opportunity to write up a blog post.

I would think that it is utterly uncontroversial that there are some fictions about real people. but, it is often the case that things I take to be utterly uncontroversial are subjects of heated philosophical dispute.

Start with this fiction (which, admittedly, has little aesthetic value):
One day, during his exile on the isle of Elba, Napoleon met a surprisingly friendly group of vampires. After a brief conversation, they parted ways and Napoleon made a mental note not to prejudge vampires. The end.

Here are a couple things I think we should all accept about that story (call it "Vampires on Elba").
1) "Vampires on Elba" is a story about Napoleon
2) The occurrences of "Napoleon" in "Vampires on Elba" refer to Napoleon.

Here is a potential worry for accepting 1 and 2:

It seems to follow from 1 and 2 that:
3) The Napoleon of "Vampires on Elba" is Napoleon.
And, intuitively we would want to accept:
4) The Napoleon of "Vampires on Elba" met some vampires.

But, now we have a contradiction, since:
5) It is not the case that Napoleon met some vampires.

So, 3, 4, and 5 form an inconsistent triad.

Let's take it for granted that 5 is beyond reproach.

We either need to deny 3, deny 4, or argue that there is equivocation going on. And if we deny 3, we will need to figure out whether to reject 1 and 2, or whether 3 doesn't really follow from them at all.

It seems to be the best strategy for rejecting 4 is to argue that it is literally false, but can be used to convey the truth:
4*) According to the fiction "Vampires on Elba", Napoleon met some vampires.

This is somewhat unsatisfying, but does allow us to cleanly preserve 3.

On the other hand, it seems to me that the best strategy for rejecting 3 involves treating "the n of S" (where 'n' is a name and 'S' is the title of a story) as a function that takes one from the referent of n to some other object. I'm going to call this approach the "Cadillac of Minivans" approach. In essence, we reject 3 because it is an identity claim, but the definite description doesn't pick out the object named by 'n', rather, it picks out some relevant counterpart of that object. Note that this is compatible with (and almost requires) treating 'n' as it occurs in 3, as retaining its ordinary reference. Just as "Cadillac" retains its ordinary reference in the phrase "The Cadillac of minivans".

I am not sure whether this strategy is compatible with accepting 1 or 2 (I am more worried about 1, I guess, than about 2).

I am currently inclined to accept 3, and to either maintain that 4 is false, but can be used to make claims about what is true according to the fiction, or maintain that 4 is ambiguous between the false reading and a reading on which it literally makes a claim about what is true according to the fiction.

I'm mostly curious as to whether there is some way to maintain natural readings of 1 and 2, while denying 3. Any thoughts?