What do you mean? I suppose there is not exact correct procedure, as it depends on the data and how they are captured.
Also, the basic steps are explained in my previous posts.
Which heights do you want to keep? As you know (I suppose), RealityCapture is working with ellipsoidal heights. Otherwise it is used more as local heights (this is not an exact explanation).
I suppose you just need import images and turn of camera priors.
Import laser scans as registered and georeferenced or not (it depends on the used coordinates and if they work for you).
Import GCPs in correct coordinate system.
Align.
Probably you will get two components (image and laser scan). There you need to place the GCPs over both components on minimally three images/LSPs for each component.
Then you will need to align your data again. These GCPs should merge the data together (I suppose they are in a wanted coordinate system, so you can set not georeferenced during the laser scan import. As you mentioned that you were not satisfied with the laser scan’s accuracy).
Also, try to aware of red triangles for GCPs.