Hi Steven

the -d argument should NOT be used for each dataset. It is there to generate and write out a default protocol (and it uses the input DWI data to do so, best use a nice/good dataset for this purpose) and then that protocol could be adapted (if so desired by the user) using the DTIPrep GUI. If you choose to adapt it in the GUI (e.g. to disable the eddy/motion correction or to change rejection thresholds), then you can also save it there (don't forget to save it).

Then you can use that protocol file for all your datasets using the "-p" option. This way you can use the same protocol file for all studies.

The reason why you should not use "-d" for each dataset is that this usage will yield to slightly different protocols for each dataset (and thus slightly different rejection thresholds) which will lead to a (slightly) inconsistent QC across the datasets. So, 
1) create a default protocol (do not run the QC itself, not necessary)
2) edit the protocol if necessary (which seems to be the case here)
3) run with -p yourprotocol.xml -c 

Martin