Long-term marginal cost is determined by the nonreusable component cost + minimum refurbish & refit costs per launch. Effectively, the benefits of reuse offer diminising marginal benefit per reuse. Though designing for long-term use (say, a personal automobile vs. a rocketship) can still make sense.
An automobile used for daily commute with a 5 year life sees ~1,000 re-uses. You can vary the parameters to get different results. There are fuel, maintanance, garaging/parking, tax/registration, and insurance costs as well.
Something New Glenn sized definitely could.
Seemingly, SpaceX is the current best in the world and they couldn’t do it.
Right now SpaceX is the only company that recovers their orbital class booster. And that didn't happen until 2016! Prior to that, everyone dismissed landing rockets as impossible. Given the lack of competition for launch vehicles + the fat defense contracts being awarded, there wasn't much incentive for the incumbents to even try to bring the cost down.
Designing a rocket that can land and be reflow is a multiyear process. As other comments have said, more design changes would been needed to F9 to make it more reusable, those changes have tradeoffs. Instead, SpaceX chose to invest in a new design and more capable vehicle.
Other launch providers like RocketLab and Blue Origin are pursuing reusable vehicles.
"Why has no one done it?" give it time... it will happen sooner than we probably expect.
Also, you’d have to redesign F9 somewhat. Starship has the fairing integrated with the upper stage, which reduces recovery costs. Starship’s booster also does Return To Launch Site nominally, which means you don’t need a separate recovery droneship (Falcon 9 does that, too, but to make the most out of every expended upper stage, it makes sense to use down-range recovery). The F9 legs are sort of complicated logistically whereas Super Heavy is supposed to be recovered by the actual launch site, reducing turnaround time and costs. Also, using methane means less coking in the engines, which reduces turnaround costs and maintenance. Lots of things like that you’d need to reengineer to enable $1 million launch costs, but it’s feasible for an MLV. Relativity is planning to do it.
As Elon likes to point out, imagine how expensive it would be to fly if, after every flight, they threw away the plane. That's exactly what people traditionally did with rockets. The result is that the cost to fly a jet is mostly the cost of the fuel, while the cost to fly a rocket is the cost of the rocket.
Consider a Falcon 9 launch. Thanks to reuse, the cost to relaunch is in the neighborhood of $30 million. The cost of the fuel is about $200,000. Over 99% the launch cost is depreciation, maintenance, or having to replace what you didn't reuse.
The aim with Starship is to make the cost of a launch as close to the cost of the fuel as they can get.
They are aiming for daily launches of starship, so the site operations costs would be divided by potentially hundreds of launches per year instead of the one or two dozen a year nowadays.
The cost model per flight of Starship is more like that of a jumbo jet than that of a Falcon 9.
20% of the cost of the whole rocket. With no reuse. Now that reuse has cut the cost of launch way down, the second stage is much more than that of the cost.