Restrict anlaysis of WGS VCF file to Refseq Select using snpeff

Question

[Gabriel]

I'm trying to annotate a human WGS VCF file to filter for biomedically relevant variants. I've run it through a pipeline using snpEff and snpSift to identify interesting variants (medium/high impact, coding, rare, etc) but when I view the variants in IGV I'm realizing many of these are to minor or crappy transcript variants, rather than the canonical one (IGV displays Refseq Select which seems similar to the "best" ones I can see in Ensembl). I've tried using the -canon filter in snpEff and it helps a little, but not much. How can I force snpEff to use the best transcripts? Ideally Refseq Select. Do I have to create a custom GRCh38 database using GFF/GTF files? Thanks