Bcftools Mpileup bias for alternative alleles?

[mcetin]

I have called snps from a bam file using the below commands:

/usr/local/sw/bcftools-1.9/bcftools mpileup -f /mnt/NEOGENE2/share/ref/genomes/hsa/hs37d5.fa -I /mnt/NEOGENE2/share/dna/hsa/comparative_seqs/lazaridis2014/Loschbour.hg19_1000g.dq.bam.hs37d5.fa.cons.90perc.bam -R /mnt/NAS/projects/2018_MCetin_Selection/imputation/Genomic_Hist_SE_Europe/new225_fixed_chr21.bed -q 30 -Q 30 -o Loschbour_ori_chr21_bcftools.bcf

/usr/local/sw/bcftools-1.9/bcftools call -mV indels Loschbour_ori_chr21_bcftools.bcf > Loschbour_ori_chr21_bcftools.vcf

Then I calculated the percentage of reference and alternative alleles that are in the vcf file. I got

0/0 10019

0/1 3412

1/1 3281

% ref= 70.1591670656

% alt= 29.8408329344

The ratio of alt alleles seem too high to me, since when I do the same thing with GATK Unified genotyper or haplotype caller, I get 5% alternative, which is I think more in line with the expectations from population genetics (Most positions should be reference)

Has Anyone seen this before?