mil. By my reckoning, that means Musk anticipates launching Falcon 5,
which has 6 Merlins on it, and presumably some other complex design
differences too, for only twice the price of Falcon I.

Does anybody see any way this makes any kind of sense?

Simplest explanation - launch costs (x per launch) and flight costs (y
per engine)

FI - $6m - $x+$y
FV - $12m - $x+$6y

12 - 6 = x - x + 6y - y
6 = 5y

so an engine costs $1.2m & therefore launch costs are $4.8m

(where "engine" = "hardware associated with an engine, inc. a share of
the rest of the rocket, assumed proportional" and "launch costs" =
upfront costs per launch, relatively constant regardless of what it is)

This is blatantly not the case, but it explains how it might work.
Nothing insane, just maths :-)

Ok. This assumes that the majority of systems used on the booster
stage will be interchangeable with the upper stage. Musk also claims
that the upper stage would be another Falcon I, meaning that he's
costing the upperstage for the Falcon V alone at the same cost as a
complete 5-engine booster stage. Possible, but again, unlikely. If you
further assume that this is reusable as Musk claims, it's going to
change your recovery systems--its a much bigger, heavier stage to
recover.

I'm not saying it's mathmatically impossible for this to be accurate,
I'm saying it doesn't appear to be likely.

(There'll be economies of scale in the tankage and structure on the
first stage, I suspect, and possibly production efficiencies on a
much greater engine-production run.)

I'll buy that for a cheaper per-pound-to-orbit cost than Falcon I. I
find it difficult to believe such a large cost jump can be made in one
step, though, unless SpaceX has taken extraordinary care and testing
to ensure that the guidance, navigation, and fuel management for
Falcon I are already forward-compatible enough to make development
costs on the Falcon V booster very very low. That would take an
extreme amount of either forsight or engineering luck.
Either way, I'll stand by until the first Falcon clears the tower.

Tom Merkle