I also don't see why you need more delta-v to go from launch orbit to circular orbit to transfer orbit compared to the transfer orbit directly. My orbital mechanics are rusty, but it seems to me that since you need to lift the perigee from Earth's surface to Mars orbit, you'll pass the circular orbit on the way.
I still don't see why circularizing the orbit would be wasteful. Circularizing an elliptical orbit at apogee still requires positive delta-v, which will help you on the way to the hyperbolic orbit. Only if you want to circularize at perigee would you need to slow down, which would indeed be extremely wasteful. It's true that you get more energy boost for the delta-v if you do it early rather than late, but I don't know if the ellipticity of the launch orbit is high enough for that difference to qualify as "extremely" wasteful. But maybe it is.
As I understand it the big drawback to EOR from an engineering standpoint is that the added complexity introduces a bunch of extra failure modes.
Of course, once you've build a big enough rocket, its usually cheaper than multiple small ones. But there is a market for medium size rockets like the Falcon Heavy already. Not so, for bigger rockets - yet.
I want to see a 140t to LEO rocket, too and I'm sure SpaceX is already building it.* However, is this really necessary? I think the orignal, soviet moon flight plan would have been to put several stages into LEO and combine them there. The plan was given up, but I think it should still be possible if no larger rocket is available soon. Consider, SpaceX plans to lunch 10 Flacon heavy (53t to LEO) per year at a combined cost of 800-1250 million.*
And of course, the Saturn V did already put 119t to LEO in 1967. Hence, their knowledge in building space stations like MIR. How much was bailing out those rust belt dinosaurs again?