Abnormal distributions in denoising QA plots

Hi Alfonso and Conn community, 

After running the denoising steps there are some scans that show a funny distribution in the voxel-to-voxel correlation histograms. I got 3 scans with a flat distribution, and another one with a bimodal distribution (see attached). I tried improving the denoising parameters by increasing the number of white matter and CSF cofounds from 5 to 10 but this did not work. What else can I try? 

Thanks in advance!

Ana Maria