You cannot take a dictionary snapshot for an entire CDB in a single flat file. You must be connected to a distinct PDB, and can take a snapshot of only that PDB in a flat file. Thus, when using a flat file dictionary, you can only mine the redo logs for the changes associated with the PDB whose data dictionary is contained within the flat file.