Hi Jingjing,
I'm sorry that you're running into some problems with yt.lagrangian_region. Let me see if I can help you.
The output of the code should return a box in code units containing all of the DM particles in the selected halo. I've only tested this with enzo outputs (and I believe that is what you're using it on too), so those code units will go from 0 to 1 in x, y, z. In addition to this output, the code also will recenter these coordinates so that the final halo coordinates will sit at 0.5, 0.5, 0.5 at the center of the box. It provides the original box shifted over in coordinates to accomodate this recentering. Lastly, you are correct that some of the coordinates will go above 1 in cases when the particles shift over the periodic boundary condition.
Do your results make more sense given this information? If you're still confused, can you send us the text of your outputs, so that we can try to figure them out?
I'm actually working on a newer version of this which will work with yt 3.0 as part of a larger effort to make efficient zoom simulations, but I've not yet finished working on it. I'll make sure to announce it here when this is complete and ready.
Cameron