MD tags with diff ref sequence

Hi, I came across this puzzling finding. I have two different reads in a bamfile, with the same exact sequence. For one of them, when I look at the MD aligned pairs, it seems like the last position (position 42) matches to the reference sequence (the T is written in capital letter case).

TTCTGAATTAGCTGTATCGTCAAGGCACTCTTGCCTACGCCAT
(42, 25398283, ‘T’)

But for the other read, the MD aligned pairs show that the last position (position 42) is a mismatch and the reference sequence should be a C at that position (C is written in small letter case).

TTCTGAATTAGCTGTATCGTCAAGGCACTCTTGCCTACGCCAT
(42, 25398283, ‘c’)

In reality, the reference sequence at that position (chr12:25398283 in hg19) is a C but I don’t understand why there is a difference in the MD aligned pairs between these two different reads that have the same exact sequence. Any ideas?

Thanks so much!

Note: this numbering is 0-based indexing in pysam. So looking up the position in a 1-based indexing genome browser would be off by 1 (25398284 instead of 25398283)