My experience in publishing (in CS) is that the publisher does not offer assistance in typesetting. Additionally reviewers are not paid for their effort, and are not affiliated with the publisher--they do it out of a sense of duty / professional advancement.
The journals offer something of value because they provide high-quality taxpayer-funded research. The publisher is nothing more than a middleman, who should be cut out of the equation.
Not sure why you're being downvoted, what you say is similar for applied math. Since a couple of years ago some publishers are asking the LaTeX source and the figures to avoid their own typesetting (save for the journal's LaTeX style).
Journals do publish a lot of work by academicians, many of them funded, in the US, by NSF grants (taxpayer money).
The journals offer something of value because they provide high-quality taxpayer-funded research. The publisher is nothing more than a middleman, who should be cut out of the equation.