|
Most probably, it's in the "convert a Word document" area. And it's not Calibre's converters, but some (really unfortunate) decisions made regarding hidden "features" in Word's storage of formatting commands — especially after changes.
Even with change-tracking off, some past formatting decisions are stored in .doc and .docx files — particularly including highlighting, which would explain the yellow bars. There are other glitches caused by things not fully documented in the file formats, especially files edited on more than one computer and/or in more than one version of Word.
Another cause can be image-file-importing issues if your Word document doesn't store the image internally (but instead stores a link to a file on the disk or even internet). I understand that post-2016 versions of Word are particularly prone to unexpected results when images are linked, but I don't use them so I haven't seen them. So, preliminarily, make sure that the images are fully embedded in the .doc/.docx file.
There are two possible solutions that will eliminate this particular problem, although they may introduce others:
1. Convert to RTF inside of Word, then use the RTF as the source. Warning: Due to the RTF file definition, this will store the equivalent of two copies of each image, so filesize will expand. If you get a warning when doing Save As inside Word that some important-to-your-book features will be lost, this option won't work (for example, headers and footers that include graphics are problems).
2. Make a clean, no-tracking-even-hidden-tracking done, version of your Word document by creating a blank Word document, then doing a Ctrl-A/Ctrl-C on the original document, switching to the new blank document, then doing Ctrl-V. Some complex formatting things may be lost (font compression in particular), and you'll need to recheck all images to make sure they transferred properly, but this will get rid of all tracked changes (including comments, another source of problems).
It's also possible that something about the text wrapping is an issue. Text wrapping should probably be reimposed freshly in the epub anyway, but that's annoyingly difficult with a lot of images (and especially if there are tables anywhere near the images). I've not run into this as an issue, but then when I work from a Word original I go through RTF anyway, especially for anything that has ever been anywhere near a Mac (it's a character-coding thing).
|