descrepancy in predicted variant count from two different versions of Mtoolbox

[dasn588]

I have observed that some human mito samples have significantly less number of variants as compared to other when using two different versions of Mtoolbox program.
For example samples from set1 analyzed with Mtoolbox-v1.1 have around 16k variants but samples from set2 analyzed with Mtoolbox-v1.2.1 have around 70-100 variants only.
Kindly let me know in detail what exactly is different in the new version of MToolBox as compared to older one which is showing such drastic variation