Also, it should be noted that mass production hits some obstacles when it comes to upper stages. You need a lot fewer engines, and higher ISP than you need for the lower stages (but not as much thrust requirement). You can do it with the same or similar ISP like SpaceX does (same engine, just vacuum optimized expansion nozzle), but that limits your scaling - it's fine to LEO/GEO but you're never going to get to Mars and back with a practical-sized rocket with those kinds of ISP figures. Which is why SpaceX's future plans hinge around in-situ methane production, so that they don't have to carry all of that return mass. It's a reasonable, although challenging, approach.
There are some possibilities mind you for getting more impulse out of their current designs. They're already taking some interest steps with the Falcon 9v1.2, aka "Full Thrust" - instead of having their LOX near its boiling point, they're supercooling it to just above its triple point and cooling the propellant to the maximum level of viscosity that their turbopumps can manage, so that they both increase in density, thus increasing both tank capacity and thrust. But while they're playing with increased viscosity propellants, they could take it to the next stage and go with mildly gelled propellants. The gelling isn't in and of itself a performance enhancer, but it lets you suspend aluminum (or if you don't mind the handling problems, lithium) particles in your fuel. Aluminum gives dozens of extra sec ISP, and lithium dozens more. Aluminum also increases propellant density, meaning more thrust and tank capacity (lithium unfortunately decreases it). While lithium metal is fairly expensive (a couple dozen dollars per kg), aluminum is cheap, about $1,50/kg.
Another nice thing (according at least to my CEA simulations with lithium) is that the latter significantly lowers chamber temperature, all other conditions (mass flow rate, expansion ratio, etc) being the same. Entering the conditions for the SSME, for example (77,5:1 expansion ratio, mass flow rate per square meter = 2223,8 kg/sec), CEA calculates (if SSME were lossless) 464,5 sec vac ISP (real world, after losses is 452 sec), 0,36g/cc propellant density, 3602,82K chamber temperature (real world 3573,15K) and exhaust of H2O (~76%) + H2 (~24%). CEA says that with a slightly different ratio you could add an extra 1,4sec ISP, but it's basically near maximum. With aluminum added to the ideal mix it calculates Al (43,9%)/LOX (39,1%)/LH2 (17,0%): 544,0 sec, 0,34g/cc, 3689,38K, -> H2 (~91%), Al2O3 (~9%). And with lithium, it calculates Li (30,0%)/LOX (34,6%)/LH2 (35,4%): 583,2 sec, 0,17g/cc, 2362,44K, -> H2 (~89%), Li2O (~11%). Now, these figures assume complete burning of the metals - which is often difficult to achieve in the real world with aluminum as its oxide has such a high melting point - but in general metalized propellants offer huge potential improvements to performance, with non-esoteric technology, and without posing serious pollution problems (like, say, using fluorine as an oxidizer does). So it'd be interesting to see what SpaceX could achieve if they could get their system to handle gelled propellants - the potential is huge.
(Note: these calculations are for adding metals to LOX/LH... but the same thing applies to hydrocarbon fuels, albeit to a slightly lesser degree)