I posted about this in another thread. But briefly :
You have RW in it’s default setting of “Links relative to page”.
When you visit the rubbish url and the 404 triggers, then the 404 page is served from this rubbish location, and links are relative to that page, as seen in a screen grab from your site below:
This is of course incorrect.
How to get around it?
-
Change your RW settings to “links are relative to site root”. (unexpected side effects may occur with some stacks and themes, test.).
-
Do the workaround I suggest in this thread :
Should we make our own 404 pages for our beautiful bespoke sites? - #18 by Doobox
If you se this workaround you should probable stop search engines indexing the 404 page as well by including this meta tag in the head of the 404 page :
<meta name="robots" content="noindex">