Registering subset point cloud to a full template point cloud

Question

I am generating point clouds which are sections of cars which contain dent information as a colour map, I need to transform these scans onto a template point cloud (i.e. the scan point cloud is of the front end of a sedan, the template point cloud is a full generalised sedan). I am looking to apply the source point cloud dent information to the correct general area on the template. Are there any pre-processing steps I can try? If registration is not what I am after, are there other techniques to achieve what I am after that I can apply.

I have been unsuccessful using both rigid and non-rigid registration techniques (FPFH + RANSAC + ICP, Feature-metric Reg, ICP + KNN refinement, TPS, etc) with a couple pre-processing steps (voxel downsample, convex-hull area scaling, bounding-box normalisation). They would get "close", but due to the differences in shape and design, the clouds tended to apply to incorrect areas of the template, or would be in the correct area with incorrect scaling and/or pose.