The scattering is given by difference in scattering length densities, between SLD of each component (e.g. lipid head or lipid tail) and the SLD of the solvent, e.g.: ΔSLDhead = SLDhead-SLDH2O.

So effectively you only have 2 SLD "degrees of freedom" (independent parameters) in the model: ΔSLD head and ΔSLD tail, that you fit with three SLD parameters.

In other words, the three SLD parameters are correlated, since you can get the same ΔSLD values with different SLD values, so the uncertainty on the derived parameter becomes infinite.

Therefore: keep SLDH2O fixed and only fit the two other SLD values.