VariantFiltration and BaseRecalibrator

So I've created a pipeline to make a known.snp.vcf for my species without known snps for BaseRecalibrator:

For a subset of my samples (~5% of my dataset, 10 samples):
I ran HaplotypeCaller on each bam file -> GenomicDBImport -> GenotypeVCFs -> SelectVariants SNP -> VariantFiltration.

My questions are:
1) is that a fair practice to estimate known SNPs?

2) Can I use that filtered vcf for BaseRecalibrator on my remaining samples? Will BaseRecalibrator recognize the PASS/FAIL and just use the PASS snps or do I need to take another step to weed out the poor quality snps?