This would be helpful to show us a screenshot of. If doing so, definitely add a colour scale bar to the FA image. Depending on the spatial resolution of your data, I’m not per se surprised that a large number of voxels get selected by that particular algorithm, in particular in an ex-vivo monkey brain (but still it would have to be supported by your spatial resolution).
Unless something that we hadn’t anticipated is wrong with your 5TT image, you may get luckier with the dhollander algorithm (if I say so myself 
), which doesn’t need a 5TT image. For ex-vivo data you may need to lower the -fa parameter, although given your highest b-value being 7000, it may not be necessary after all. See a few posts about this over here:
- Estimate response function for multi-shell ex-vivo marmoset data - #10 by ThijsDhollander
- Multi-tissue CSD
But definitely screen shot those FA and 5TT images for us… who knows what we might see.