DQXI Migration to DS - CandidateSelection transfrom differences

Question

The migration from the DQXI CandidateSelection transform to Data Services Match transform version continues to be a thorn in my side.
The client uses DQXI in real-time.  They string together multiple CandidateSelection transforms before submitting the candidate list to match.  In one of these CandidateSelection transforms, they have a collection set and each record in the collection set results in a new record appended to the collection.  For example, the incoming record generated 3 candidates (4 records in total) in an earlier transform and then in this downstream CandidateSelection transform, each of those 4 records generates an additional candidate for a total now of 8 records.
When I migrated these to Data Services, only the first record in the "collection" is being used to generate candidates.  It appears as though this is a result of the option EXECUTE_SQL_PER > ALL_RECORDS_IN_COLLECTION no longer existing.
I absolutely require this functionality in Data Services.  Is there a suggested work around?  Keep in mind that the current DQXI query references 3 different source tables and that this is a real-time dataflow. 
I've pasted the statement below for reference.
Thanks,
Chuck
SELECT *
FROM
DS_PRS_EXT_PUB PUB,
DS_PRS_EXT_PUB2 PUB2,
DS_PRS_BREAKCODE BRCD
WHERE
PUB.PRS_ID= [TEMP_PRS_ID] AND
PUB.PRS_ID=PUB2.PRS_ID(+) AND
PUB.PRS_ID=BRCD.PRS_ID(+) AND
PUB.PRS_ACTIVE_IND = 1

Answer 1

Chuck,
The CandSel fires SQL for each break group. Since there is no break key defined, it is assuming all the records are part of same break group. One way to solve this would be to define a "pseudo-break key" which is unique for each row, then it should work.
By the way, I think you should explore ways to do this with Query join etc. I can't tell exactly how without knowing the whole schema, but Cand Sel in this case would end up doing very similar thing underneath anyway.
Abhiram