From what I've read operator skill is a huge variable with RFA, so a 20% variable is reasonable.
A very unfortunate skip of Phase 2, as they could have had an efficacy measure and further refined the resulting Phase 3.
Next big mistake as you say is not having a long enough study to determine the OS from the micro-metastisis which may be the true value of the Thermodox platform. It could be statistically not justified for the PFS endpoint, but a huge benefit for the OS.
Overconfidence is such an easy thing to get misled by. In trading and research.