Well done.
If you look closely you will see that the time scale from the paper's graphs were slightly different on the combined graph than on the broken out graphs. Take a look at those 2 events just prior to 36 months. They are a few days earlier on the combined chart that on the un-meth chart.
That is the reason your combined chart looks a bit farther off.
On the meth chart, I think there might be an extra censor around month 21 causing the subsequent drift in the Y axis.
BTW, I thought your work post 36 months was very interesting, albeit speculative. The math looked right. That was a good catch that there must be an event at those specific times. A few other notes:
. Dr B's chart from ASCO had a few more events post 36 months on it (the data was the same, the time axis was clipped a tad further out).
. There could be an event past 88.2, no way to know that.
. I do not take that number 6 as known to be exact. It varies with the number of LTFU's prior to 36 months.