Nice work overall!
I haven't taken a close look at it yet, but PyGlossary (based on your GH profile and the output, it appears that that's what you're using to convert the dictionaries) has some bugs in it's handling of prefixes for Kobo's dictionary format.
In particular, uppercase words (e.g. Cumberland in sv-en) will appear in the autocomplete, but will appear as no definition found. See
here for my notes on proper prefix generation.
One way to fix this automatically without making modifications to PyGlossary would be to use dictzip-decompile and dictgen (see my thread about dictutil) to regenerate the dictionaries.