I struggled with this for a while. Here is what I discovered:
If you have a duplication that fails, the next time your duplication runs it trys to get that nights back up plus the one that failed the night before (so it trys to dup two nights worth of backup - a good idea on Veritas' part assuming you have enough tape space to dup two+ nights of data). If you have a very small retention window for your backup-to-disk media set such that it gets overwritten right-away, the duplication job cannot find the media for the previous night(s) and hence fails with the error you are describing.
This is normally not a problem because most people keep more than a few days of backup-to-desk. But if your dup job fails two or three nights is a row (say your tape operator is on vacation or sick and there is no one on site to pop in a new tape each day) , it keeps trying to back up all the sets that is has missed (2 then 3 then 4 etc.) and eventually hits a day where the backup-to-disk has already be overwritten/erased and generates this error.
I have been told Veritas is going to fix this in a future release. In the mean time, I have found deleting both the back-to-disk and dup jobs and then recreating them clears out this counter and tells the dup job to backup just that nights backup.
Clear as mud?