Since the pool size changes a lot, 'her' ramp down will not be smooth, and in this case there was a big gap of no blocks as you can see.
From 2019‑02‑04 to 2019‑02‑22 was a 339.455% block.
So it ramped down 3.4Nd with no blocks until the end of that 3.4Nd ramp down.
However, 'she' basically stopped mining before that as you can see the ramp down already happening from the 1st of Feb
(from 126.27THs on 17-Jan)
So by the time 22-Feb came along, 'she' had already almost completed ramping down.
The only reason it wasn't completely ramped down was due to the two small spikes up from zero and back down again you can see on 'her' Shift Graph on the 7th and 12th of Feb.
Unfortunately the only fix to that would have been to not switch away from the pool during the bad luck and be ramped down when the good luck arrived.
All the numbers on PPLNS are 'expected' not, 'will happen'.
You 'expect' on average, to have a block every 100%.
Instead what 'will happen' is blocks will be all over the place.
The 5Nd ramp down 'expects', on average, 5 blocks, one every 100%
But that's not what will happen of course, it will be random and in some rare cases there will be none for 5Nd (any block 500% or more)
Bottom line, 'she' simply chose the wrong time to leave the pool.
You can't know in advance when will be a good or bad time to leave, but alas that was probably the worst time to leave due to missing out on the good luck also.