But the 28.2% is almost a final figure. If it showed a confidence interval it would be no more than 1% either way. Only a handful of subjects who joined late Oct/Nov would not have their data included. Perhaps 10 or less. Given the indicators of ridiculously good survival for the 2015 intake, I would forecast that this 28.2% will not go down. rather it is likely to go up a fraction of 1%.
Of course the bigger and more important question is what is the separation between treatment and control on OS36.
It still went up from 2017, and your good point about the censors is likely moot, or rather a “wash” due to mixed bag. I’d say 28.2 is final +/- half a point.