I think that water spasing defined by some forces (like Van der Vaals). And this size is more o less constant in one lipid membrabe. If we speak about small angle scattering (X-ray or neutron) we must remember that all sizes, which we obtain by this method, are averaged over all sample volume. If you have one water spacing at one part of vesicle, another spacing at another part - SAXS will give you some average size from all such parts which are in X-ray beam.

Dmytro is correct, the balance of forces (van der Waals, hydration, undulation repulsion, electrostatic) is responsible for the water spacing and SAXS gives an average of water spacings. Papers by Horia Petrache, Adrian Parsegian, John Nagle and Stephanie Tristram-Nagle might be helpful.

It isn't so much a matter of technique used to determine spacing* as it is about the state of equilibrium between different forces to achieve the least strained lipid layer. 

* Averaging of X-ray diffraction can easily be overcome by actually viewing individual segments of lipid bilayers using electron microscopy and such measurements have been made decades ago.

It depends on the amount of water, since the multilamellar system forms in excess but not too much of water. Otherwise you will get micelles or vesicles or even monolayer at the surface (Langmuir–Blodgett thin films). The thing is that SAXS is not so sensitive to that variation of multilamellar water thickness. But the main reason why lipids are at the surface and with more or less the same water thickness between their bilayers is a density effect and the hydration force (after fully hydrating lipid heads water is expelled from the gap due to the combination of gravity-density and electrostatic-hydrophobic forces, but with the increase of water amount that gap due to Brownian motion and entropy gradient will increase, however SAXS cannot efficiently trace that amount of increase due to lack of sensitivity-resolution and perfect samples).