SUGGESTION: to add an option to produce 3D models from pre-derived point clouds by creating triangualr faces from three closest vertices in the point cloud, i.e without extrapolating or approximating more complicated faces out of wider range of vertices.
REASON: with advancement in drone positioning (e.g. Mavic M3M with RTK module), images from UAVs can be now geolocated significantly more precisely, even to single centimeters. This implies craetion of point clouds with points much more accurately oriented and positioned in 3D space, implying that approximating faces out of a larger group of points may not be needed anymore. Actually from my practice this proves to work even worse in such situations, as I have noticed when processing data from M3M in PIX4DMapper. The suggested (simplified) approach is likely to produce 3D models much faster and also likely to avoid strange artifacts, especially out of very dense point clouds.
OPTION: an option could be added to consider only point within a specific region (distance) for such a simplified trriangular face generation, while using a standard approach in case that points are more spaced away from one another that a pre-defined limit.