IMHO they aren't all variants
HarperColins is a Parent company
Harper is an Imprint of HarperColins
and Harper Colins is just wrongly done
My personal preference is to put the IMPRINT in the publishers column, rather that the Mega Publishing conglomerate .
Usually the Imprint is found on the publishers Title page, or the copyright page
It is also unfortunate that even the publishers can't be consistant.
Baen Books
Baen Publishing Enterprises
Sometimes the imprint breaks it down to Baen SF, Baen Fantasy
Unfortunately, some of the metadata sources are comiled from various user inputs, creating even more variants.
Manual TLC is the only way