Hello thans,
Actually, dense cloud is generated according to the extrinsic and intrinsic camera orientation estimated during alignment/optimization, so those parameters would not be refined on the further steps.
Access to dense cloud - chunk.denseCloud, to mesh: chunk.model (if == None - no dense cloud or mesh).
You can check the number of projections for valid tie points on each image, it there are only few (like 10-20, for example), you can reset camera alignment.
Also you can check the estimated camera coordinates, if you have any criterion how to check if it is close enough to the object.