What Adjust said.
Plus: epub (well, OCF to be precise) specifies for two files to be be present with specific names: mimetype and META-INF/container.xml. The first one declares the zip to be of type epub, the second one serves as starting point telling the program the location and type of the content of the book.
With the space in the directory the file did not have the required name anymore - therefore to the epub engine the the second file (and thus the information on how to find the book content) was missing.
You're right in that windows will automatically erase tailing spaces from created directories (which isn't true for neither Linux nor OSX - didn't know that last tidbit myself until I unpacked the ebook in question).
Unfortunately the zipping process is not simply copying files from A to B. You can declare the specific content entry in the zip header (think file system in a single file) while creating your ebook - and my guess is that something went wrong in that step.
And the Moral of the Story: Computers suck. Use Magic Marker on Papyrus for a change... 