mne_inverse_operator

Hi MNE users,

I am trying to use the source covariance matrix option --srccov in
mne_inverse_operator but I am getting the following error:

"...
Creating the source covariance matrix:
       Loaded a priori source weighting from
/homes/14/asalazar/anes07/structural_mri/eeg_mri/simu/eeganes07_simu_sigma2V_source_cov.fif.
Cannot handle nondiagonal source covariance matrix yet
..."

I am working with matlab simulated data and have my own source covariance
matrix. I found a way to save my source cov. matrix using
mne_write_cov_file but since I need its tree-like structure (which I do
not have), I loaded a noise covariance matrix and changed some of the
elements in the structure (kind =2, dim = # sources, names = [], proj =
[], data = the calculated source cov), then I saved the file using
mne_write_cov_file.

The source covariance matrix is a 20454X20454 diagonal matrix but
mne_inverse_operator keeps complaining the matrix is not diagonal.

I know the problem could be the tree structure I used for the source
covariance matrix. Could anyone share with me how to save it properly?
Also, what could be the reason for mne_inverse_operator giving me this
non-diagonal error?

Help is really appreciated!!

Thank you very much!