Results are in, but they're not good. Bottom line is that I can't fix the problem.

I have tested the two files you sent me (one containing a "Thumbs.db" file, the other containing only images). Ubooquity failed to open both of them on my PC (Windows 7).
I uncompressed and recompressed both of them with Winrar. After that, both were succesfully imported in Ubooquity.

I really don't think that the "thumbs.db" file is the cause of the problem, as the archive opening fails long before Ubooquity tries to list the contained files. The error I get comes directly from the external library I use to read rar files (same problem Emmanuel had).

The problem is that this library is not maintained anymore and I did not find any alternative library (compression libraries are not the most fun or easy things to write). So there is nothing I can do about it, sorry.

Your options: recompress the files that are not preoperly read. Or convert them to cbz (that's what I do as I don't like using RAR files, they are not an open standard so there are much less tools/libraries to work with them).

Great idea !
I'll do that.

Ubooquity does not care (in theory) about what kind of files are in the archive. It just keeps the ones it identifies as being images (jpeg, png, gif and, in a future version, webp).

If you can, could send me one or a few files that still can't be opened by Ubooquity after recompression ?
Email to: tom ' at ' vaemendis ' dot ' net

That great news ! :D
Let me know if you encounter any difficulty coming from the way I implemented the OPDS feed.
I tried to stick to the specifications as closely as possible but I found some parts quite ambiguous.

Could you take a look at your log file and check if you have some errors ?

For now I'm quite puzzled by this problem. So if you have time to "play" a little bit with the settings (try to share another folder, create another user etc.) and find some hints, it would help a lot.

Ubooquity never modifies the user's files (books or comics). That's a strong design principle that will not change.
So modifying comics metadata is not an option.

However, keeping the unred/read state of a given book in the database (as well as the current page number for comics read using the online reader) is something that can be done (and already requested a few times).
So why not.