Focal shift correction
In the presence of a Refractive Index Mismatch the axial movement of the microscope stage is not followed by the focus. This causes a geometrical distortion of the acquired dataset. See the Fishtank Effect.
The Huygens Software considers the refractive index mismatch to adapt the calculated theoretical Point Spread Function (PSF) when Doing Deconvolution, but does not correct this geometrical distortion afterwards. This can be done simply by manually changing the Sampling Distance later. Do not do it before deconvolution when using a Theoretical Psf, because its calculation will take the mismatch into account and distort it accordingly. If the image is corrected but the PSF is not, the deconvolution will return bad results.
A more difficult case happens when you use an air lens corrected to work as a, let's say, glycerol lens. See Air Lens Correction for details.
When using an Experimental Psf there is no need to modify it, because if it was measured using the same conditions the distortion is also present there. What matters at the end is that both the PSF and the image have the same distortion. Correcting the image geometry afterwards can be seen as simply a matter of data representation. In this case, though, the correction can be applied before deconvolution to both the image and the PSF, by simply setting the Z step to the real value.
Other problems
Remind that on top of this geometrical distortion there is also the effect of the Spherical Aberration. Its effect is to elongate the PSF intrinsically, and therefore you have two "overlapped" elongations. The intrinsic elongation will also affect the image in a way that deconvolution will manage to reduce, but it is always a better idea to minimize the mismatch during acquisition, e.g. using a water lens instead of an oil one. More details in Mismatch Distorts Psf.