Sounds like the book was one of the fortunately-rare Topaz books.
They are essentially a series of images with OCR data, and when DeDRMed they become HTMLZ files instead of the more usual AZW3/MOBI.
There is literally nothing you can do about them -- they never had the data to begin with, but you thought they did because of the pictures.
If you take a look at the raw HTML, you will probably find that there is no actual Mathjax there -- not to mention, Kindles don't actually support Mathjax to begin with, so that was never your problem...