Dear Marlene,
I would suggest you to first coregister your image to the SPM template before running freesurfer. Then the easiest thing to do would be to use freesurfer meshes for all boundaries and then you won't have a problem. You could do that in SPM12 batch tool for head model generation. Also if your only problem is that your mesh 'sticks out' of the inner skull boundary, that shouldn't be a problem in up-to-date SPM12 because there is code there that automatically makes sure that the mesh is far enough from the boundary for the forward computation to be valid. This was not done in SPM8 though. Look carefully at the final display of the forward model and there you should see that he mesh is within the boundary, even of originally it wasn't.
Best,
Vladimir