The Data Collection guidelines that are agreed to at the time of upload are there for a reason and state that high amounts of overlap should be used. 75% should be considered the minimum because it provides the system with 16 looks at everything in the scene.
With only 20% overlap the majority of the area is only getting one look while some smaller areas are getting 2 or 4 looks. This is not nearly enough to be able to reconstruct the scene.
You are not looking for a photogrammetry based solution, you are looking for a image projection or ground intersection based solution which is something we do not currently offer publicly.