Vikas,
I recently did a simulation with a setup similar to what you described. In my simulation, I had water passing through a carbon nanotube pore. I applied pressure to squeeze the water through the pore using a reflective wall. I applied pressure by changing the position of the wall with a variable command. I set the initial position of top wall to a value approximately equal to sigma for the LJ parameter for O-O. I think I used about 3.5 angstroms (not nm). I figured this little bit of vacuum space at the top of the box would let the system equilibrate a little better. I worried that if the H2O molecules were to close to the reflective wall initially something bad might happen. As the simulation ran, the water expanded a little and filled up the vacuum space.
I don’t think it would matter if you set your reflective wall position to one value and your zhi value to another as long as zhi>the wall position. The kspace slab option shouldn’t allow interactions across periodic images in the z-direction anyway. Seems like setting zhi and your wall position to the same value would help ensure the best distribution of atoms across processors and might help speed up your simulation. Maybe someone else could confirm this.
Good luck.