This message can happen for two reasons:
1) the b-values of non-diffusion-weighted measurements are recorded not as 0 in the bval file.
This often happens with the b-values for these measurements are not exactly 0, instead they take a small value such as 5.
From your email, it looks like this isn't your problem. But if it is, you will not get any sensible NODDI maps. To fix this problem, one can call FSL2Procotol with one extra argument
protocol = FSL2Protocol(bval, bvec, b0threshold);
2) for some of the voxels, none of their b=0 measurements is greater than 0.
This often happens for the voxels at the edge of binary brain mask. If this is the case, you will be able to get sensible NODDI maps for the other voxels. So you can safely ignore this error message.