|
Looking at your files in Writer, the images are anchored "to Paragraph" in the odt and "to Character" in the docx. I'm not sure how the anchoring in Writer translates during conversion, but you could try adjusting that.
Personally, I never try and do anything much with images in Writer, because I know they will almost never look just the same (or even close!) in an epub (or azw3) after conversion. I just insert them where they need to be, anchored "as character" since that seems to generate less messy code. Then I use the Editor with the converted book to code the images as I actually want them to appear. And that seems to involve some black magic, since a given image and it's coding that I make look great on my Kobo, will often fail after a conversion to azw3 for my Kindle, and I have to edit it again. Or vice-versa.
The bottom line is, Calibre always makes some assumptions in converting. You never get exactly the look of the word processor file. I used to go crazy messing with the conversion options, and ended up just learning how to edit the things. It's easier in the long run.
Or look into a product like Jutoh, a desktop ebook publishing tool meant more for professional output. Jutoh has an "almost WYSIWYG" interface, and when you use it to compile an epub, for example, it delivers a very faithful rendition.
|