Typesetting is done by authors (in some fields) and then completely redone by Elsevier staff. The value they add is questionable at best, but they certainly spend a lot of effort making the paper conform to their standards.
...or detract, one might add. I know of a researcher who has been meticulous about editing and wording, only to have the submissions "corrected" for publication to use the wrong word (existing, and meaning something different) for key concepts.
I'm guessing incompetence combined with standard dictionary computer spellchecking at the publisher's end.