What benefit would they have with their own, incompatible with anything else, platform?

This way they can build differentiation full Android stack integrators can't. That, of course, if QNX proves a better environment than Linux.

This is, BTW, the same strategy Microsoft adopted eons ago, when the de-facto standard business OS was CP/M. It was very easy to port your programs from CP/M to MS-DOS (thanks, in no small part, to Intel making the 8086 asm-source compatible with the 8085) and that allowed them to have a reasonable software base from the start. Unfortunately, for CP/M, this was a one way road - it was not as easy to port back.

If Rim can manage a superset of Android, they may have a fighting chance to differentiate themselves.

I fully understand the advantages of having your own platform.

At this point, I'm more concerned about their competency than their strategy.

I don't see having an exclusive platform as an advantage right now. If they could pull off iOS API compatibility, that would be brilliant (assuming their patent cross licensing with Apple adequately covered their asses) and could be conceivably be done starting with GNUStep, but Android seems a good second-best option.

OTOH, I believe the Apple app store TOS would prohibit any in-store app to be distributed in any way other than the store, so, there is one interesting contractual risk.