I found the wiring diagram from here: https://arcarc.xmission.com/PDF_Arcade_Atari_Kee/Asteroids_Cocktail/Asteroids%20Cocktail_DP-150-1st-01A.pdf
Upon further examination, it looks like you're
at least partly correct, and
probably completely correct.
1. The video inverter theory of operation on
sheet 2B indicates that J20 pins 7 and 19 change between +5 VDC and ground, depending on which player's video is being displayed.
In a cocktail game, the wiring harness shorts connector J20's output pin 7 input pin 19. When the video of player 1 is being displayed, pins 7 and 19 are +5 VDC. This results in a non-inverted video output. When the video for player 2 is being displayed, pins 7 and 19 are grounded.
2. J20 pins 7 and 19 are also connected to P24B pin 6 (player 2's controls), which appears to be wired the same as P24A (player 1's controls), except for pin 6.
3. Pin 6 on P24A provides the ground for player 1's buttons, so it appears that pin 6 on P24B provides the ground for player 2's buttons. (unconfirmed, but strongly implied by the schematic on sheet 1A)
4. Player Input Circuitry theory of operation on sheet 2B indicates that switch inputs are active when pulled to ground.
Applying this info to two cocktail cab game-play scenarios:
Scenario A - P2 presses a button during P1's turn.
- J20 Pins 7 and 19 are at 5v ==> P24B pin 6 is at 5v ==> pressing the button applies 5v to the input ==> input is not pulled to ground ==> button press is not registered.
Scenario B - P2 presses a button during P2's turn.
- J20 Pins 7 and 19 are grounded ==> P24B pin 6 is grounded ==> pressing the button applies ground to the input ==> input is pulled to ground ==> button press is registered.
The only other thing not yet confirmed is if J20 pin J is always in the opposite state as J20 pin 7 when the PCB is in cocktail mode. (i.e. when one is grounded, the other is at 5v)
- If anyone has an Asteroids cocktail cabinet and can confirm/deny that J20 pin J works like this, it could help make MAME more accurate. Confirming that there's no continuity between J20 pin J and the ground plane of any Asteroids PCB would also work since it would confirm that J20 pin J is not hard-wired to ground. (same PCB is used for upright and cocktail cabs)
- If it does operate like that, MAME should define separate P2 inputs for cocktail cabinet configuration.
I had mapped the controls just like you pointed out, but the downside is that players can interfere (on purpose or accidentally) with each other's game.
True, but until we can confirm the theory I layed out above and find a MAME dev to update the Asteroids driver accordingly, double-mapping is the only way to make Asteroids work with P2 controls.

Scott