Hello Bryan,
The representation of the object as point clouds is so that users can have an idea of which part of the model would be a better target to track and if the object is actually a good target. However, we do realize that positioning augmentations in such a preview is pretty tricky. The only solution would be simply trial and error. We do have it in our feature list to replace the point clouds with the actual 3D model but this is not something that will happen soon.
Thanks
Eva
Bflynn882