Using media queries and breakpoints is a good solution that does make the code a little hard to read. I've seen some css shorthand which might simplify things. MIGHT.
I had some interesting issues with my image-heavy Soldier Boys ebooks. I have several other ebooks down the road which will use images extensively; that's why we really need better support for images on Kindle and other systems.
I would never expect pixel-perfect rendering for ebook these days, but I think the person reading expects graphics to be rendered reasonably well (and floated if possible).
Kindle Unpack doesn't need Python. I don't have Python on my computer and wouldn't know how to use it if I did! It's been a couple years since I used Kindle Unpack, but at that time it yielded up both a KF7 and a KF8 book, complete. Now it probably also yields up a KFX (Kindle Format Ten) book. And I understand that if you upload a Word doc, you also see your original document.
I don't use media queries. I usually go for a width="100%" instruction on images, and KF 7 does that by default. The result is good. I have two books in which I used either a "sort" at a 12 percent width or mug shots at 20 percent, and those I finessed by uploading a gif or a jpeg physically small enough that the image as reproduced looked okay.
I also have a colophon for my publishing imprint, so I upload my entire title page as a gif file, which IMHO looks more professional anyhow, and is just what most Big Five publishers do.
(Don't trust KDP to publish a print edition. Don't trust CreateSpace to publish an ebook.)
I also have a plugin that will convert any html doctype to a full epub. I haven't released this Sigil plugin to the general public on MR yet because I think it might be contentious -- might release it in the near future. This plugin should convert a valid html doc from DocBook directly to full epub format. And even though it will not format floating images, this plugin should do everything else for you including create stylesheets, split the html into xhtml files, create a valid TOC and Nav TOC etc. In other words, this plugin should help to save you a heck of a lot of time in your own manual conversions using DocBook or HTMLBook.
If you want to try the plugin -- ConvertHTML2Epub -- then just get my email from my KDP profile and drop me line and I will be happy send you the plugin. And if you have any problems with it -- just let me know. Oh...and if you use this plugin then, after running it, you must remember to delete the original DocBook HTML file from Sigil's Book Browser otherwise it will cause Epubcheck errors. Other than that, this plugin is very easy to use.
I don't have Python on my computer and wouldn't know how to use it if I did!
You don't have to install Python now on your op system to run a Sigil plugin because Sigil already comes with its own internal or bundled Python(since v9.0 I think.). Just tick the Use Bundled Python option in Sigil's Plugins > Manage Plugins and you should be able to run most, if not all of the plugins available now in the Sigil Plugin Index(including all my plugins).