Generally PSA is used to load the source system data as it is, with the same structure and meta data. We filter things out or use transformations when we load the PSA data into DSOs or InfoCubes.
The source systems have a limited downtime, they cannot be kept out of service for long. So, the data to be loaded is copied as it is to the PSA.
If you do not need the 3 fields in the DataTarget, I suggest you skip them while loading data to the target (InfoProvider) i.e when you create the mappings.
As far as filtering at PSA level is concerned, I will try that out. Currently, I dont have access to any BI system. I will let you know, once I have tried. Even I am new to BI.