Kai,
10µA output current is enough to drive darlington

Well, do you see any reason to use such scheme (here: micro-current with darlington input)? Maybe for mobile usage with battery power? Well, but how much does MCU eat itself when works? (=
From one side, micro-current schematic requires good filters and other protections against parasite signals and noise. From other side, typical darlington transistor provides gain about 1000, so 10µA in base produces 10ma at collector. Nice, why not use ordinary transistor with 10k in base then? It does the same result but with standard and stable way.
BTW, here there is another note against usage of darlington transistor directly with port pin. The problem is that such transistors has high value of base-emitter saturation voltage. Depend both on base and collector current in may change at 1...2,5V or worse. As we know, pFET3 switches on/off when pin voltage is about 1,4...1,5V. So we have undefined state here: depend on type of transistor and its collector current, it may be that pFET3 is at ON or OFF state, so base current cannot be guaranteed and all the device we construct is not stable.
Okay, I agree with you that RMW-instructions are elaborated for correct work with ports` bits. I never say against this fact of matter. But I still cannot see a reason to connect base of transistor to port pin directly (=

Regards,
Oleg