I like your model. But for me the efficacy question would really boild down to whether the PFS is compelling or not, so I would like to do this:
a) Place an 80% probability on senario 2a)ODAC finds the PFS data compelling. If this happens, the probability of accelerated approval is 90%
b) Place a 20% probability on senario 2b)ODS does not find the PFS data compelling. If this happens probability of accelerated approval is 10% (higher than your 4% - due to historical precedence such as ovarian and panitumumab)
Total probabiliy of accelerated approval 0.8*0.9+0.2*0.1=0.74.