We’ve had very good results from pix4d testing using very dense photo sets. If there aren’t quite enough photos, it tends to produce several “chunks” of model, but then dumps them in a single point cloud and tries to mesh them.
Having played with VisualSFM, I presume this is the same situation that leads it to produce multiple point clouds. Is that right? That seems a more useful approach, as then I can examine the different parts more easily (which would make it easier to assess how to try to fix the problem), and might even be able to use one or more of them.
Is there any way to separate these chunks in pix4d, and to densify and mesh one or other of them?