Firstlevel design matrix efficiency in FSL looks wrong when using fmriprep output

Hi all,

My firstlevel design matrix in FSL consists of 12 EVs + temporal derivatives, and the following confounds: GSR, 6 motion transforms and motion outliers estimated with FD. I used space-MNI152NLin6Asym_desc-preproc_bold.nii.gz as input and derived the confounds from the confounds_regressors.tsv files from the fmrioutput. The firstlevel analysis ran without errors and the output looks good, except for my covariance matrix and design efficiency. My covariance matrix has a very bright spot, due to the similarity between rot_x and rot_y as regressors and my design efficiency only shows zeroes apart from the first eigen value. Does someone know what the reason is, if it’s a big problem and if so, how I can fix this?