This is a known bug but we haven’t managed to fix it yet. Note that changing the last part (G/F/WGS) does work as expected. (That link has an explanation of why it happens if you’re interested, as well as why it doesn’t happen for our core datasets such as /ncov/....)