This is probably better suited as an issue on Github, but in the current implementation, the approach and retreat trajectories are calculated via computeCartesianPath, which does not use the max velocity factor. The trajectory has to be time-parametrized manually. This is set to be streamlined in the future, but if you need it in the meantime, you could patch it roughly like this:

1. Find the lines where the pick/place functions execute the plans (this should be it for picking)

2. Parametrize the trajectories with your desired scaling factor. There are explanations here and here, among other places.

3. Let the pipeline execute the time-parametrized trajectories.

You can open a pull request or issue here if you decide to pursue this. It should not be too complicated.