It's fantastic that you've started to work on bigendian support :-)
So, as I wrote yesterday and the ntfs-3g README file explains, the expected
behavior is exactly the same what you experience on bigendian ARM. Since
the next ARM problems are in ntfs-3g thus please let's restrict the
discussion only to the ntfs-devel list.
The very good news is that that this is quite easy to fix, it just needs a
dedicated hardware and plenty free time which you seem to perfectly match.
In the first step, please just make sure that the relevant on-disk metadata
accesses are little endian (most of the code is already ok), then we could
continue with the time consuming code validation process. Thanks!