You're referring to primary outcome measure, not primary...

You're referring to primary outcome measure, not primary endpoint. The endpoint is the point at which a defined number of patients have met a condition (in this case progression). An endpoint can be defined by something other than the the outcome measure.

The reason why I said "highly likely" is that it again comes down to probability distribution. "Normal" FOLFIRI efficacy is around 4-4.5 mths, but has been less than 4 in at least one trial. It seems highly UNLIKELY for a population to jump like we saw in the HA-I trial without there being a contributing factor making them different to previous trials.

6 weeks median PFS advantage is not the defined primary outcome measure. It is just median PFS. The 6 weeks is the difference assumed when they sized the number of trial patients to give a statistically significant measure of difference. The actual statistical significance depends on the trial measurements, not the assumption. So you could get a lower measured difference with less certainty. Or if the distribution of outcomes was tighter than expected you could get a lower difference with adequate certainty.

I'd imagine they would have given themselves some fat when designing the trial in case there was more data spread than guessed.

The reason why big pharma have big expensive trials is so they can get drugs approved with minimal efficacy gain. The large size of the trial gives them sufficient statistical certainty for small differences