RMAN Duplicate database taking longer each day (going back to old archivelogs)

We have an 11.1.0.7 RMAN duplicate database job that runs daily, both db's on same server.  Normally the aux db recovery only requires the last archvie log from the target db.  Thus recovery is fast.  For the last 3 days it has been progressively taking longer, always going back to start at the archivelog required 4 days ago for the last "good" duplicate.  e.g last good duplicate only required log 33125.  The next day also started with that log, so recovering up to the current archivelog.  Then the next day still goes back to that 33125 log and recovers through 2