You should calculate from Falcon 9 Block 3, which will have 50% more payload to LEO (15 tons). I just saw the numbers for that earlier and those 5 tons extra could make a world of a difference, especially since the Dragon capsule is not that heavy and Falcon 9 does have lots of savety margins on top of that.
I guess that if they need to use the margins due to a failure of some sort, they will simply write that stage off as a loss and dump it into the ocean. If they dont have to use them, then they do already have some extra fuel left in the tanks. Add in the 5 tons of extra payload and trade them for reusability and you might actually get somewhere. I think that at least first stage reuse does move into the area of being quite possible with that. I am not so sure about the second stage, but the first stage is the more expensive part anyway (with 9 Merlin engines and lots of structure).
The second stage only a has a single engine and less materials. So the loss of that would be more acceptable. From what I see SpaceX planing for the next few years, it seems that they will first try only reusing the first stage and then the second stage.
The last comment on NBF got me thinking--how much of the stage really needs to be recovered? If the engines themselves are 60% of the cost, then just recovering them would save you massive amounts. Not sure if that would be easier, but might be better than trying to recover the whole thing.
I think that Musk also wants to reduce turn arround time. If you can just take the first stage as it is and add a new stack on top, then fly again a few days later, that does make a difference. I also want to point out that Musk is always thinking in the long term. He wants to colonize other planets. You need cheap and responsive access to space for that.
