My first thought on seeing your results is that you’re probably using lmax 6 in the dwi2fod call (the default with your data). Trying setting it to -lmax 8, see if that helps.
Also consider using multi-tissue CSD if you’re not already – it can be used with ‘single-shell’ data, since the b=0 volumes constitute a ‘shell’ in their own right… See e.g. this thread for example.