Samtools mpileup and bcftools only call for SNPs and Indels and don't do any filtering on the variants. The "FILTER" column is part of the VCF file and various filtering tools such as snpEff, GATK VariantFilteration and VCF tools will use the "FILTER" column as a flag to tell if a particular variant passed filtering criteria provided by the user. Normally for SNPs, you at least require 5 reads with unique start positions that confirm alternate allele. Some people use 3 reads. Other criteria like mapping quality, Strand Bias etc. are used too. Read VCF tools. In my opinion, none of the SNPs above are high quality calls. The mapping qualities are good but number of supporting reads are too less. Also, supporting reads were only aligned to one strand.

The low score in the QUAL column (and the QC entry) is a red flag, as is the low depth. Total depth can be seen in the DP value, the depth when low quality reads are filtered out is in the DP4 info; that's all explained in the vcf standards, which, if they aren't explained in the headers of your vcf, can be found by googling.

Based on that data lone, you probably should not conclude anything at all about what the sequence is there.

I think you should further proceed using annovar. It's an annotation tool and also helps in filtering based on quality, depth etc. The two snp's given by you are of poor quality. But you might have more snp's which may be of good quality.