Question: What determines which archive log file the capture process will start on

After stopping and restarting the Streams Capture process, in the alert log for the source database, I see that the Logminer has started processing an archive log file created prior to the one it was processing when the process was stopped.  Sometimes, the archive log file it restarts on is the one that we force at midnight each night to include the Data Dictionary for the database, but sometimes the archive log it starts on is just a few sequences back from the one it was processing when it was stopped and none of those archive log files include the data dictionary.  Is there a