I placed many control points on every slice in the brainmask.mgz file.
Then I re-ran recon-all using -autorecon2-cp argument. But the white matter volume did not grow into those regions with control points. The process completed without any errors.
Any idea on how I can include those regions in white matter?
I also tried to use 'mri_segment' for lowering the white matter low limit to 60. That did include some parts of the missing lobes, but also included grey matter from frontal and parietal lobes. So, that approach was not ideal.