COLMAP is generally underrated IMO. It's what I primarily use because it's very flexible for different pipelines, it produces one of the best SFM results as a starting place, and it's open source with direct access to the database and decent documentation.
