I think the point is that "assume" is a weird word: it is probably true, in a statistical sense, that married couples are more valuable to some companies, and less to others, and their models reflect that.

No one is saying "eh, screw the single people/happy couples" and tweaking the scores in some ad hoc way.