the probability of success in second indications—when PoC had presumably been established before any clinical trials—was lower on average than the probability of success in primary indications,
I suspect that second indications might include more investigator-sponsored trials, and also the bar for proceeding to a pivotal trial might be lower when the drug is already approved. Further, the very low success rate in difficult-to-treat cancers (particularly for approved drugs) might simply reflect desperation - let's try already approved drug X in case it shows some efficacy.
Bottom line, I wouldn't give the results of a database dump like this a lot of credibility.