less % coverage of metaG after adding more references to database

[AnneliektH]

Ran sourmash gather on pig gut metagenomes against gtdbtk. Assembled MAGs from the metagenomes, then reran gather against gtdbtk + newly assembled MAGs. For most of the metagenomes, total % coverage of a metagenome (sum of f_unique_weighted), goes up after adding these newly assembled MAGs to the databases used. But for three of them, the total % coverage goes down. Why is that?