Hi Arkiev,
Given you’re on 3.0.2 and not 3.0.3, it’s possibly due to inconsequential differences in “diffusion densitisation direction” for b=0 volumes, which was reported here and fixed here. Is the third volume in the input data a b=0 volume?
If it’s not that, then I would need to see the full gradient table of the concatenated DWI series.
Cheers
Rob