Let me answer my own question - I figured it out.
In Curry, create ‘surface points’. I used a 10mm spacing which creates ~1400 points. Then import points to localize. Then export localize to matlab which creates a work space called ‘curryloc’.
You need to invert the matrix:
curryloc_alt = curryloc’
Then save this as an ASCII:
save (‘curryloc.txt’, ‘curryloc_alt’, ‘-ascii’)
This file can then be imported into Brainstorm as head points. It will query the need to scale and suggest 0.1.
Then refine using head points as usual.