I hope that what you write to with DataStage is not a data warehouse, or I have hunch you have some architectural problems. In that case, run to the library/bookstore and get all Kimball books
For your current problem, the problem is that you can't process data which is not there.
The standard way to do stuff like that is sandboxing: you keep a copy of the data, compare everything on a new run and make rows for record deletions/inserts/updates to be applied to the real data. (also described in the Kimball books e.g.)
Alternatively you could change your business process to give you actual input when sales reps leave the company (rows you can use to delete them).
The second alternative would be better as it would minimize "guessing", e.g. once upon a time you only get half an input file for sales reps and your job decides to close off half of your sales reps.
Ogmios
In theory there's no difference between theory and practice. In practice there is.