Me again, with another question about HTML5 output, mostly unrelated to my question about strange headers, hence the new post.
By exporting our HTML5 directly from our ditamap, we were able to circumvent the header problem (where the chapter output template would grab a weird header and then use it as a title for every section of the HTML5 output). Since it gives us the option to apply a ditaval in the publish panel, I figure this is an intended workflow.
Now with graphics, however, the publish operation does not convert our .ai files to .png. It keeps the path to our original .ai files and they don't display.
Why won't it convert the graphics? Any way to get it to do so? Do I have to use svg graphics if I want to keep using vector formats for my graphics so they'll work in browsers? I see that there is an "SVG to raster" option in the settings, but I'm not using SVG in the first place.