No concrete solution but a few suggestions and comments that may provide a lead.
Firstly I'm not convinced that Retropin is on the correct track. It is certainly true that you need to be careful when replacing 74 series chips with substitutes. In essence you can always use a plain 74?? chip, you can also always replace one of these with a 74S, which you don't see many off in my experience. Trying to use a 74LS or 74L will most likely cause problems, but a 74LS can replace a 74L. But if the input is being detected in switch diagnostics then I'd say all is ok with the buffer chip. I read Retropin's comments on the difference WMS used in test mode (and I'd be interested to hear more details on this) but in test mode I can't see how you can frig this. The game will scan the switch matrix in the same way as in game play, the only difference is that the program output is directed to drive the displays rather than solenoids etc.
Next thought is when you say you've tested the cpu with Leon test rom then I'd question how thoroughly you did these tests. If you whip all the game roms out, install the test rom, power up and watch the leds and then push the diagnostic button to watch the leds for expected outcome then that's good and necessary for success but you need more. With the tests just as above you aren't examining the addressing of the game and flipper roms. I'd stick all the chips you plan to use in along with the test chip and then use a logic probe (my preference) or scope to examine each pin on every chip for activity. You certainly should look at address, data, chip select, R/W and VMA on each chip. My fairly strong guess is that you'll see a missing signal somewhere here. Looking at the picture on the thread with the super cap installation around the 5101 the traces look a bit ratty
Finally if this is the same board you've done the super cap mod on I'd be inclined to disconnect this. I had a quick look over the circuit and reckon what you've done is perfectly ok, but I remove it just to confirm.
