Does Harvester’s de-dup routine use all fields/metadata to calculate the MD5 hash for files?

For loose files, only the full file data is used. Filesystem data such as the filename and the MAC times are not used to calculate the hash of a file.