Krall mentioned Plushenko's 4T leap so it's possible the science even breaks down different parts of the jump from different models. Not sure, just guessing. All his own jumps are recorded and analysed too so I'm sure his own best 4T is studied for replication also. (At this point I think his 4T surpasses Plush's because of his more intricate entries.)
All the models also offer data for extremely detailed micro analysis so it's not necessary that he learns by simply copying one specific model. Of course, if he should copy one model, I'm sure he copies the best there is.
I am pretty sure there are now skaters who have Chan's quads on their learning software too.