Survival rates are nearly always expressed as Medians, it's a...

Survival rates are nearly always expressed as Medians, it's a "stats thing" as Means can be ill defined, because what is the X-axis? number of people or time (days, months)? The distribution of the X-axis then may not be normal. Median has other benefits, it can be quicker to gauge the trial outcome as not all patients need to be tested, and it is more intuitive to explain to patients.

I am not sure why it was inferred that the trial was destined to fail? Trials maybe conceived with a degree of emotion ("want to succeed!") but are modelled and tested objectively (statistical modelling, power analysis) before patient testing.

"if ALL the liver-only patients showed more than 3.1 months PFS improvement, the overall PFS for the group would still be >3.1"
Probably missing your point here, but that will be correct for mean or median, because in this instance you have defined 100% of your group, it is statistically impossible for the group to have <3.1.