As with anything else it depends what you optimise for and at the most basic level the car uses the engines to accelerate at all speeds but might need to use break pads at slow speeds.

I've seen data showing ~70% recovery on a Model 3, but I also can't find the source :/

Like OP mentioned, the efficiency of recovery varies by speed but is generally in the 65%-80% range for total round-trip (the system captures ~80% of the kinetic energy but with round-trip losses, you only get ~80% of that energy to propel the car).

Maybe you're thinking of the total range boost provided by regen braking? That's often shown to be roughly 20% via studies;

https://sci-hub.se/10.1109/vppc.2011.6043109

> Simulations show that the energy reduction of the vehicles under test can be more than 20% by applying regenerative braking.

It's an older study, but they show the regen efficiency at 60% for the EcoTruck and 70% for the bus.

It is basic physic.

There is all the time friction of air and the road. You can never recover that.

And in rare cases where you actually can recover energy, at non emergency slowing down, it is probably indeed in that ballpark, wich is something(especially in city with stop and go), but not very high or much. With very high tech, you can increase that number a bit, but not worth the effort. Electric cars need good batteries and every improvement there is good.

It depends a lot on how you drive. If you do a lot of hard braking, the car will have to use a lot of friction braking. If you slow down more gradually, the car can just use regen which is gonna be pretty efficient.

Still, I'm not sure OP's train of thought holds. Mass matters because you need to spend energy to accelerate it and move it uphill. Regenerative breaking does not have 100% efficiency. Therefore, energy losses are proportional to weight. Also, isn't tire drag proportional to weight? That's a energy sink that can't be recovered by breaking.