Am not sure how you have designed the jobs. Assuming following jobs / table details :

Table : EMP_TAB
Job1: Hash file extraction from EMP_TAB ( with key details, e.g. month date if its a monthly load)
Job 2: Cleanse/Transform ( Apply all business logic)
Job 3: Load job with look up on hash file from Job1 (to EMP_TAB)

Decription:
The extract job should be executed first in the sequence. For all the jobs check the option "Do no checkpoint run" except the first hash creation job.

Now in case your source had 1000 rows, and the job failed on 200th row. On simpley restarting the sequence the hash file would have all the rows inserted by earlier run, & the earlier jobs like job 2 would not run unecessarily. The load job would ensure that where the match is found , not action is performed.

Thanks
Pranav