there is no point to look at the whole G+ population in ph1/2 as predicative for ph2b because ph2b include only significant G+ pts where majority will be triple negative.
I don't see the logic here - but I am not sure it matters unless you are saying that you expect the ORR in the treated group to be closer to 20% than 10%??? (I don't think you are, in fact, saying this - but because I don't understand your logic I am unsure.)