Should, and may, the panel of normals only include germline variants present in two or more normal samples?

Should the panel of normals include or exclude common SNPs, or does it not matter?

Should the panel of normals include or exclude variants suspected of being artifacts, or does it not matter?

I include everything present in 3 normals, without any filtering.  

Th3 likelihood model will skip all variants that have > max.pon (runAbsoluteCN, default 3) hits in the PoN and are either annotated as somatic (with a matched normal) or NOT as a common SNP (tumor-only mode). These will still show up in the output, but will be flagged.