Do we need to recover_dynamics again when we redraw velocity map for differential kinetic regimes?

[hyjforesight]

Hello scVelo,
In the Recompute velocities step of the Differential Kinetics tutorial (https://scvelo.readthedocs.io/DifferentialKinetics/#), you use scv.tl.velocity() and scv.tl.velocity_graph() without running scv.tl.recover_dynamics() in advance.

I run scv.tl.recover_dynamics() twice. First time is before the differential kinetic test and top-likelihood genes testing. Second time is after the differential kinetic test and top-likelihood genes testing. But the number of recover genes are different (955 vs 794). See below.

Does it mean that Differential Kinetic Test and top-likelihood genes test remove some genes? Can I skip these 2 steps and just add diff_kinetics=True into this fuction scv.tl.velocity(adata, mode='dynamical'), if I know clusters are absolutely in different kinetic regimes?
Thanks!
Best,
YJ