To me, saying WYSIWYM is ultimately no different than WYSIWYG other than a single letter change. The development of WYMEditor seems to be generally headed in the direction of WYSIWYG - making this no better than TinyMCE. I'd rather see the editor take the direction it originally started in where "What You Mean" to do reigns supreme.
IMO, block-level elements are generally broken in every browser's idea of 'designmode'. The 'p' tag within WYMEditor, however, seems VERY stable across browsers. So, in theory, it should be possible to use the 'p' tag for every block-level element and then just apply visual styles based on classes. It should be possible to use jQuery to parse the data while it is being pulled for block-level transformations (e.g. a 'p' tag with a 'h1' class becomes a 'h1' tag in the output) and to transform tags as they are going into the editor (e.g. a 'h1' tag becomes a 'p' tag with a 'h1' class). This approach would allow you to eliminate all block-level elements except the stable 'p' tag and the result would be a very stable editor that is focused on What You Mean instead of What You See.