Is there a magical time boundary for diagnosing eyewitness identification accuracy in sequential lineups?