One interesting counterpoint on costs that is often made (But I think is incorrect), is that with a reusable booster, you make less of them. Which means you are less efficient, and you need to design a line for building them slowly, in a cost effective manner.
This is an argument Arianespace has made.
SpaceX may well face this issue. If they can use the booster 10 times, then they really only need 2 boosters a year to fulfill their manifest.
But they have a manufacturing line designed to build 10-15 cores with sufficient engines (100-150) a year. If you run that at 2-3 a year, you will waste a lot of money and the cost of the booster will go up.
But then SpaceX seems to be cleverer than your average bear, and you may have noticed an odd (for booster builders) decision they made, to keep the second stage structure, the same as the first stage.
Often the structures taper/different diameters, or different fuels, or different materials. In the Falcon 9 case, the first stage and second stage use the same fuels, materials, engines (for the most part M1D vac vs M1D are close enough for this argument).
Thus SpaceX can beat this arguement by ramping up second stage production to keep the line active at an efficient rate.
Engine wise, they still have to slow down, since 1 vs 9 engines is a big rate change for the line). But happily the Raptor engine is much the same size, and needs much the same tooling to build, and they are going to need to make crap loads of them. (31 on the BFB, and 7 on the BFS, so each full ship set will need 38 engines!)
So there are many tradeoffs, and clearly SpaceX believes one way, and prior to their proving it can work, everyone else in the industry believed (and had their own numerical models to prove it) the other way.