After inserting several image tags into a document, the Ebook process picks up some images correctly, but reuses others. The PDF compile process works correctly.
I’ve attached a small project based on your General Non-Fiction template. It contains a short manuscript with three images referenced via image tags.
To reproduce the problem, compile the manuscript using the default PDF and Kindle KF8/mobi formats. The PDF file should show three different images. The Kindle version should show image temp-1, temp_2, and then temp_1 again (even though the third image tag references temp3).
This problem took a while to troubleshoot because we have several images in each chapter, and each image name started with chaptername_ followed by a unique identifier. For .mobi output, every image in a chapter ended up being the same. I saw the comment on p. 420 of the manual ("If for some reason you have two graphics named…); that was my first hint toward identifying the issue, even though we have no fully-duplicate image names (only the first part of the names are the same). Maybe the image name match for an Ebook is less thorough than the match for PDF.
p.s. When generating several iterations of .mobi output with the same filename, it looked like some of the previous files were not cleaned up/removed; old images showed up. This made it hard to tell if my latest changes worked. This might be a Kindle issue (not L&L), but thought I’d mention it. My workaround was to iterate the output filename on every compile.
test-images.scriv.zip (288 KB)