For arguments sake, lets assume that the second stage was 10% of the manufacturing cost and the first stage was 90%. You could argue, that the engines are what's most expensive and the first stage has nine times as many. Ignoring the cost for lunch, landing, refuelling etc. you could at best reduce the cost by 90%, a single order of magnitude. Unfortunately, the first stage might be worth much more. I wouldn't be surprised if the cost distribution is closer to 50-50. The second stage has all the avionics, a special vacuum version of the Merlin engine and might be more weight optimised due to the larger effect on maximum payload.
Lets hope the Mars rocket (BFR) will allow full, low cost reusability. I'm sure SpaceX is working very hard on that.
Also, while expensive, plane changes are still cheaper delta-V-wise than punching through the soup our atmosphere is.
Also, changing orbits is expensive. Remember you need to change both height and angle, and the delta-V is huge.
Bringing them back down is almost certainly the easiest option.
Engineers have been scratching their heads for decades trying to figure out where all the money goes. For shuttle launches, depending on who's doing the accounting, fuel was something on the order of 2% of total launch costs. And the shuttle was reusable as well.
I think it's that disparity that attracted Musk to the rocket business in the first place. You look at the cost breakdowns and think "What? How can this be so expensive?" If everything goes his way he'll have cut the vehicle cost by a factor of at least 20 by the time he's done.
I could see SpaceX segmenting their customers into two tiers: mission completion sensitive, and cost sensitive, with who gets which rockets straightforward.