As you have mentioned, you have different occurrences for below mentioned fields:

PER_START_DT (20 occurrences)
WL_INCM_RATE_PC (6 occurrences)

You can not project them together in the same row set (i.e. in the same branch downstream of the Normalizer Transformation). If you do so, you will get the same error "Normalizer norm:: Ni-or-1 rule violated".
Normalizer transformation expects fields should have same occurrences that you want to flatten under one downstream passive transformation (like expression transformation).

How Normalizer transformation work when you have different set of occurrences for different fields ?

At this moment it works like a router. It creates its own group depending upon number of fields with different occurrences. And each group will behave actively as it can have different numbers of records depending upon the occurrences, hence you cannot merge then into one single passive transformation (Expression transformation).
You can connect different exp. transformation to the same Normalizer to create different branch for each field(group) having different occurrences. In your case there should be 2 exp. trans after the normalizer. You have to project only one specific filed with occurrence in one expression transformation, Use two instances of the same target connected to these 2 different exp. trans.

So, depending upon the number of fields with different occurrences, that many branch should be created after Normalizer. Then either you can merge these branches using some active transformation (UNION etc) or you can connect to different target instances depending upon your need.

Let me know if you still have some doubts or in case of any concerns..

Regards,
Shadab