If you could brake with two wheels until the point of slippage, then yes, you wouldn't get more than double. But that's not the case. Braking with just two wheels completely unbalances the car. It makes turning trajectories to go awry when braking. It puts the car at a risk of going into a spin after two tires hit a poodle or drive over a manhole cover. If you are braking with just two wheels, you can't get to any considerable braking power without getting a completely unbalanced car. Accelerating is a different matter, because when you accelerate most of the weight goes to the rear wheels, but when you are braking most of the weight goes to the front wheels, so the unbalancing effect is seriously increased.
Agreed that the limit being applied by traction control will need to be much less than the ultimate point of slippage (and much less than the limit applied when accelerating). But whatever that limit is to give a comfortable margin of safety when the driver just 'lets go', I can't see that it can be any higher (for the same margin) once the front wheels start braking too - if anything, it's going to be worse because the braking at the front further unloads the rear.
On the other hand, I was ignoring the fact that weight transfer means the maximum braking at the front (to an equivalent margin of traction) will be greater than the rear, so I admit my analysis was too simplistic and on a purely traction-limited basis there could be more than twice the regen braking available on four wheels compared to just the rear two.
However, the real question is how much regen is in fact limited by traction. My suspicion is that the 60kW top limit comes mainly from the battery, but there might still be useful savings where the current system doesn't actually reach the top limit - in the slope up to full power as the regen comes on, and/or at lower speeds.
BTW, I just did a different experiment today which confirms my calculations to some extent. I have a steep, long incline on the way home. I coasted down with the pedal up, and the car sped and then stayed at 25mph, at about 40KW regen power. On the way back, I drove up at 25mph constant speed. The car needed more than 80kw (couldn't do a good measurement since it was just one run, but I think it was close to 100Kw) to maintain that speed when going up. So there is indeed a very significant difference between the energy going in and going out for the whole cycle. Even discounting drag and friction, and considering that the discharge cycle is more than 90% efficient, I would be surprised if the regen cycle was much more than 50% efficient today.
That's a good experiment, eliminating more of the variables, and certainly confirms as we suspected that the round-trip efficiency of the drive-regen-drive cycle is not very high. But I'm suspicious of your 90% number: it seems high for the overall drive case. Maybe it's a figure for just one element in the drive chain?
For the battery, it is difficult to clearly allocate losses to discharge vs charge, since every use of the battery involves both: I suspect it's common to regard drive efficiency as being from the output of the battery to the wheels, with all losses in the battery being 'charging losses' (between what you put in to the charger and what you get out, the two things that are easily measured - which lumps together losses which occurred during charging plus those that occurred during discharge). So our round-trip regen has regen losses (motor/electrical from wheels to battery input), battery losses, and drive losses from battery output back to wheels.
I would expect the losses in the motor to be symmetric - though this immediately gives you a lower efficiency when measured as a percentage. If you have a particular combination of rotor/stator currents that gives you best efficiency when driving at a certain torque (and hence output power at constant speed), then the same currents in reverse will give you the same regen torque. The losses in the motor are all also controlled by those currents, so you will have the same numerical losses in the regen case, but you are comparing them against a smaller number so the percentage efficiency is less. For example, if the motor is 90% efficient and you are driving 100kW, then 90kW is going to the output shaft and 10kW in losses: with regen at the exact same torque and speed, there is 90kW coming in through the shaft, the exact same 10kW going to losses, so only 80kW electrical output, so only 88% efficiency.
The drive electronics is harder to reason about without knowledge of the topology Tesla have chosen. Efficiency here is mainly a cost/weight trade-off (unlike the motor which has more constraints), and the drive/regen paths could be shared or separate to varying degrees, so it is possible that Tesla has sacrificed some efficiency in the regen side, though I would expect this to be the most efficient step of the whole chain and so not a big number either way.
These losses do add up though: if we guess at 88% motor efficiency in regen,. 95% electronics efficiency (each direction), 80% battery efficiency (charge/discharge combined), and 90% motor efficiency when driving, that gives:
88% * 95% * 80% * 95% * 90% = 57% round trip efficiency
You seem to be measuring something a bit worse than this, but it's not far off. So while there's always improvement to be made, I'm not convinced there's huge 'low hanging fruit' here to be picked.
[disclaimer: my experience of building motor drives is with BLDC rather than induction, and at lower powers, though I believe the physics is largely the same].