Line AP crosses the incircle in two points as do lines BP and CP. The applet picks the points nearest the vertices. However if X', Y', Z' are the farthest from the vertices points, then, say DX, EY' and FZ' are also concurrent as are the triplets DX', EY and FZ' and DX', EY' and FZ.
If P lies on the incircle, the result is trivial. So assume not.
Since the result deals only with tangency and concurrency, and there is a projective transformation fixing the incircle of ABC and taking P to the incentre, we may without loss of generality take P to be the incentre of ABC. But then by symmetry X lies at the midpoint of the arc EF so DX is the internal angle bisector of EDF. Similar statments hold for EY and FZ, so these three lines are concurrent at the incentre of DEF, as required.
Since DX', EY', FZ' are external bisectors of ΔEDF the triplets (DX, EY', FZ'), (DX', EY, FZ') and (DX', EY', FZ) are also concurrent, the triplet (DX', EY', FZ') is not.