I agree this forced closed box mesh can be a bit of a pain, and its not about not having enough coverage in the photos. I have attached a screen grab of something I did last week that required me to spend some time cleaning up the mesh to remove all the extra faces generated. (aside from generating all this extra mesh that takes time to clean it also wastes a huge amount of texture space).
So for this example we used a drone to take +400 images of a rocky hill top. the scan processed excellently. But on export I get all these sidewalls and flat bottom (essentially the reconstruction region) added to the model that I have to go an cut away in a third party application. I would rather just get the mesh as its displayed by the points as it would save a lot of time.
I hope that make sense.