I've had a look at these images and the problem with registering these masks is twofold:

 1 - The centre of the FOV and the coordinate origins are very different and when these are aligned (as is done for the initial guess) there is zero overlap of the masks, and this causes the registration to fail.  The solution to this is to align the Centre Of Gravity instead.  I've written a schedule file (attached) that does this instead.

 2 - One mask is much bigger than the other and so there are lots of positions where one mask is 100% within the other.  Hence the registration method does not know how to prefer one over another, and so it will just find any of these 100% options.  There is no particularly good solution to this except to make the masks more similar in terms of size.  It does make me wonder if the voxel resolutions in the images are set correctly.  So I'd check that.

Hope this helps.

All the best,

Mark