Paircounts problems using new version

Hello
I have the old version of CUTE working in a local cluster, I moved to the new version for using it in NERSC and the paircounts computed with the new version disagree with old version for the same parameters and xis make no sense, do you have any idea where this could come from? 
Mariana