Discrepancies in analysis

While analyzing the same dataset with Metaphlan2 and Metaphlan3, there are huge discrepancies in the microbiome profile with less than 50% similarity between any 2 samples. How is the change supposed to be interpreted?

I would say it is mainly because of the greatly expanded database of bacterial genomes used. The pre-print in bioRxiv states “MetaPhlAn 3 incorporates 13500 species (more than twice than
MetaPhlAn 2) with a completely new set of 1.1 million marker genes.” If the marker genes changed a lot, I expect that results would change a lot, too.