Compiled Image formatting problems

While preparing a Kindle book, I want to share the manuscript with my editor(s).
I have 31 Figures as images.

So I want to be able to quickly compile to several formats (epub/kindle, docx & pdf).
I have used a horizontal resolution of 1440 pixels for my images so that they appear at maximum quality on iPads, tablets etc…
I set the image size to 15.8 cm which should enable them to sit almost perfectly on an A4 page with 1 inch margins.
I format to PNG as GIFs seem to lose the image size data on compile.

The Good news is that when I embed the images (via links) for my figures, they appear fine in epub/mobi if I use Compile “As Is” for each image folder. This at least gives me hope for the ebook.

However, when I compile for PDF, the images are too large and overflow the page sides considerably. The only solution I have found so far is to reduce the pixel size of all the images to make them smaller. I know I can do this tempoarily but would prefer to find an intelligent solution. Am I missing something obvious?

When I compile for docx, all the images fit the page, but most have been resized to 55%, 62%, 67%, 71%, 79% etc. but why these different values?
Also I have no idea why just 2 out of my 31 images should appear fine at 100%, very mysterious.

In the compiled Word document I can select each image, Then click: Format, Size, Reset, to get back to the intended full size, but I dont want to do this for all those undersiszed images every time I send an updated manuscript out.

Can anyone recommend a Scrivener book that deals with compiling images to any useful extent? I have several that don’t.


As for my final Kindle publication:
Having read around this further, I now realise that I need all my Kindle ebook images to be either GIFs or jpgs (with qual <50) and file size less than 127kB (though 124kB to be sure).
These are the necessary conditions to prevent Kindle messing with my images at all.

So ideally Scrivener would give me the option to compile images in these formats (GIF and JPG) without messing with them (so they pass through the process transparently)

So as things stand at present, I guess I have to go through an intermediate such as Sigil.