For velocity you're using the velocity of the actuator, instead of hte derivative of the position error, right?
For the standard: "look how well it's working!" step response test, there is no difference. But once you have a dynamic target, that doesn't hold up, Wouldn't it be better to take the difference of the target-velocyt and the actuator velocity?
Simone giertz had a "slave robot" as her latest video. I'm under time-pressure right now, so here is a link to one of her videos. Find her latest video yourself... https://www.youtube.com/watch?v=WcW70-6eQcY
In the case of dynamic targets, like with her slave robots, also taking the average of the target positions speed would help get the same response to a step as with a "step plus linear movement".