For a project that requires you to measure many points, process the images progressively.
First, mark around 20-40 key points. These key points should be well spread over the object volume. Then setup reference points, constraint, and initial imaging parameters. Invoke 3D computation, using only the TR or TR(F) mode. Do not try to use TRF or TRFK even if F and K are unknown. Use guessed values for F instead. Examine the result of the computation and resolve any problems.
After initial iteration, mark more points, 20-50 at a time, and re-invoke 3D computation. With more point marked, and possibly more reference points and constraints, more imaging parameters can be computed for better results. Examine the result after each computation.
For a large project, the iterative process makes it much easier to discover errors in marking, reference points, constraint, and initial imaging parameters.