One of the most significant points in the Moon's favor is the (relative) ease with which a space elevator could be built. It doesn't need unobtainium, current structural materials are strong enough.
Anyone who plays KSP can confirm this for you, especially using the Real Solar System mod, though the same lessson is apparent even in the stock game.
Is that 2.1km/s really a whole extra stage? Or can it be met by reducing cargo capacity somewhat?
Two interesting points about Red Dragon - it wasn't feasible to put parachutes on it so its aerobraking would have been greatly limited. It wouldn't have much surface to use for deceleration in that case. I'm curious just how long it would have spent aerobraking, SpaceX's docs don't seem to mention that. And it was cancelled a few months ago with the intent of designing a "vastly" heavier lander, for which aerobraking will be that much less practical.
edit: But I can do a back of the envelope calculation. Decelerating 6500kg from 6km/s to 540m/s is 116G joules. The diameter is 3.6 meters so a molecule of air in from of the Dragon would be displaced (on average) 0.53 meters. Let's assume the atmosphere is a constant density (0.020 kg/m^3). Assume a drag coefficient of 0.05 (and ignoring the lifting body effect) that is around 200kN of force, tapering down to 1.5kN. That is like 3000 km of aerobraking at maximum atmospheric density and should take around 2300 seconds. But that depends a lot on the deorbit profile. I'm going to need a bigger simulation...
It's not an ideal design, but this way they would only have had to make minor changes to the capsule, mainly to add more fuel for the landing than the regular design.
I believe the extra delta-V you need to get to Mars is less than the delta-V you need to slow down to land on the Moon.