GATK HTC snp events at adjacent bases

I concurrently used haplotypecaller and samtools mpileup for calling phase events on exome data. In certain edge-cases I found that if the SNPs are on adjacent residues, HTC treats them as two separate calls instead of one. I found this behavior consistently in a bunch of different exomes I tested. Can anyone suggest me if this is expected or such cases should be treated specially depending on exonic position, allele frequency or strandedness etc and treat either as a single call or separate based on that?
for example : chr3:52392018 G>T and chr3:52392019 G>T.