The entrance pupil, stop and exit pupil are all images of each other, and so any of the three can be used as the STOP surface in the LDE.
My preference is always to make the real limiting aperture the STOP surface. In this case that would be the grating. If there is any optical power between the STOP surface and the first surface of the system, turn on ray-aiming.
Sometimes people prefer to make the entrance pupil the STOP surface. The benefit is that this is the fastest definition to launch rays with. However on modern machines this is not such a big benefit as it once was. The downside is pupil aberration. You would typically make the STOP at the entrance pupil larger than it needs to be, hence overfilling the pupil, in order for rays to correctly illuminate the real limiting aperture.
That’s why I prefer to set the real limiting aperture as the STOP and turn on ray-aiming.
-Mark