We compare several basic embodiments of a recently proposed and demonstrated micrometer-scale Spatially Offset Raman Spectroscopy (micro-SORS). Micro-SORS is a recently introduced analytical method for noninvasive characterisation of the chemical composition of subsurface, micrometre-scale-thick diffusely scattering layers at depths beyond the reach of conventional confocal Raman microscopy. The technique is applicable, for example, in nondestructive subsurface analysis of highly scattering stratified matrices such as painted layers in cultural heritage or in noninvasive analysis of stratified polymer systems or biological samples. Using Monte Carlo simulations, we analysed two defocusing variants of micro-SORS and a variant involving a full separation of illumination and collection zones on the sample surface. Both the penetration depth into the sample and relative enhancement of sublayer Raman signals were studied as a function of layer thickness and type of technique and their parameters. The model predicts that the most effective method by far is the variant with fully spatially separated illumination and collection zones. On the other hand, the defocusing micro-SORS, where both the laser and Raman collection zones are defocussed and overlapped, yielded the lowest performance although its key benefit lies in its simplicity as the concept can be practiced on existing conventional Raman microscopes without any modifications. A basic experimental verification of the theoretical findings contrasting two extreme modalities, the fully separated micro-SORS variant with the defocusing one, is also presented.