Hi Luca, Probably it is not clear from the documentation of mergeDWIs, but mergeDWIs does not perform any type of registration or alignment across DWIs. It literally just concatenates one DWI volume after another and is written for a different use case (where diffusion dataset from same subject is somehow split into multiple files, either by the scanner or because of the way data was acquired).
From your last answer, it seems that you are interested in studying the group average diffusion properties. Is that correct? If so, then you would need to co-register the subjects to the template and then transfer the subject data to the template. This can be done in most straight-forward fashion if you first run BDP in the native space (i.e. run BDP separately for each subject in usual fashion) which will give you diffusion properties like FA, MD etc in the native-space of each subject. Then you can use affine/non-registration to transfer these diffusion properties to the template and look at statistics like mean, variance etc. If you want use BrainSuite template then running SVReg on your subject’s T1w data would the recommended way. If you want to use ICBM template then you would need to register each of the subject to the template and transfer FA/MD/others (using register_files_affine or other tool of you choice).