I think your question is actually about why publishers don't make the LaTeX source of papers available to readers rather than why publishers don't accept submissions in LaTeX form. You might want to clarify your question.
Some publishers prefer to accept PDF versions of paper for review, but then ask for LaTeX source code after the paper has been accepted. Doing the peer review process with a PDF version of the paper saves the publisher the trouble of running the paper through LaTeX and fixing any problems that the authors might inadvertently have introduced into the manuscript (such as using non-standard packages of macros.)
At the final publication stage, authors typically submit LaTeX source to journals. The journal then applies its own style to the paper, adds copyright notices and page numbers, and produces a final version of the paper using LaTeX. However, journals typically only publish PDF versions of the papers rather than the LaTeX source.
Many journals make their style files available to authors and ask them to prepare the manuscript using the journal's style. This helps to avoid problems when the final version of the paper is prepared by the publisher.
Having the LaTeX source of a paper makes it slightly easier for plagiarists to cut and paste mathematical formulas and text from the paper or to maliciously produce alternate versions of the paper. Commercial publishers are also generally opposed to any use of a paper that goes beyond simply reading the paper- making LaTeX source available tends to make such reuse easier.