That they are not Lattice doesn't matter. I thought perhaps the same-- the chip on PowerCache was an MMI. I bought new old stock MMI and AMD PAL16L8 and programmed them without the security option, and they both could be read back easily. It was fairly common practice to set the security bit by default during manufacture, so really not too surprising.
For the 16L8, there is still chance the Charles MacDonald adapter can read the chip, if the logic doesn't involve latches or flip-flops, or you'll get a partial decode of the equations along with a bunch of garbage on the latch pins output in the form of an .oe equation.
Do recognize that The Rejuvenator replaces the existing daughterboard. Most likely, the contents of the PALs are going to be very close to the original Commodore versions, which we know the equations
I would think the best bet would be to trace the pins of each PAL and identify which signals are on which pins. Then reference the original Commodore equations and create an approximation. Then use a logic analyzer comparing the original Rejuvenator PALs with the approximation could give a very good idea of what would need fixing. Be sure your replacement chips are the same speed or faster than the existing.
Another option I have been toying with is using an Arduino Mega to send every possible combination of inputs to the chip while monitoring the outputs and create a truth table which could then be minimized to the proper equations. It's tough to realize, however. The logic can get exponentially difficult to decipher once latches, flip-flops, and output enable equations get assigned and looped back into other equations. I think the 20L8 only allows for 5 or 6 types of latches, and there are so few product terms that it just might be possible. I'm currently working on this to decode some other 20L8, I have not had any success.
Of course, if the logic is elementary, then a simple breadboard with some switches and LEDs could be used to fabricate the truth table.