The Gadgeel trial had a median survival of 34 months - which is impressive - but only had a sample size of 28. It also was a single-arm study, and the regimen being studied (Alimta+cisplatin+docetaxel) is investigational. So it might not be reasonable to assume that patients receiving standard of care therapies would have the same median survival.
The Chang study had a slightly larger sample size (n=44), but it too is a single-arm study where patients were receiving an experimental treatment which conceivably could increase median survival time compared to standard of care.
The Franca study, if I have the right one, is not a randomized trial but a retrospective study. One of the selection criteria was to include patients who underwent surgery with curative intent - so these patients would be expected to have a better prognosis than usual. Furthermore, it only included 27 Stage III patients.
The Kelly trial was also mentioned by the previous SA article, and this is a better trial to use to estimate what the median survival time will be in the START trial. It was a multi-centre trial and the data are based on 243 patients. However, the selection criteria for this study were rather stringent and only about half of the patients who were initially eligible underwent randomization as you can see in this chart. So essentially the trial weeded out patients who were likely to have a worse prognosis.
Market Data 
Markets 
AI
