If a content spec uses a locale other than english and includes characters other than basic ASCII characters than the build will fail because the books entity filename isn't URL encoded.
This fails in publican with an error about not being able to find Book_Info.xml and appears to be a bug as based on the spec and running xmllint. Though that needs more investigation and also some relevant test cases.
I've come up with a workaround which is to change the mainfile name to "Book" or "Article" based on the type (this is only done when the title contains those invalid characters). This means that the entity file name doesn't contain any characters that would need to be encoded.
The particular project I needed this for is moved out of the PressGang system, thus the fix is no longer needed.
Moving the status to CLOSED.