HaplotypeCaller missing a clear variant

Dear colleagues,

I am comparing two callers and found some cases in which HC behaves strangely.
Let's look into one of the cases. There's a site in which there are 15 bases overlapping. Of these, 6 are like the reference (C) and 9 are variants (T). Please, look at the attached IGV's snapshot. The reads are of good quality; same for the bases. And there are no indels nor mismatches around. How is it possible that HC genotypes this site as 0/0?

20 60216651 . C . . . AN=2;DP=14 GT:AD:DP:RGQ 0/0:6:14:0

These are the running parameters:
-T HaplotypeCaller --genotyping_mode DISCOVERY -ERC BP_RESOLUTION -R human.fasta --dbsnp dbsnp_138.vcf.gz --ploidy 2 -o output.gvcf.gz -I input.bam

Then:
-T GenotypeGVCFs -R human.fasta --includeNonVariantSites --dbsnp dbsnp_138.vcf.gz --variant output.gvcf.gz -o output.genotypes.gvcf.gz

I am using GATK version v3.4-0-g7e26428

Thanks!
Federico