Inconsistency between vcf and m3vcf reference panels

[eyherabh]

I've found that, for chromosome 1 in the 1000 Genomes Phase 3 Version 5 reference panel, the VCF contains two entries for 15274725:T:<CN0> and 227955266:TTG:T, and for each variant, their corresponding two entries in the VCF do not have the same genotypes for all samples. Are these differences are a bug or is there a suitable explanation? Thank you very much.