NEI can be thought of as just replacing the OSC2 waveform with the output of the previous track. So instead of a simple square or saw, OSC2 will spit out whatever modulated mess you‘ve already patched on the track before. This, in essences, allows you to apply a whole other set of filters, modulations, etc. to further shape the sound.
This is extremely powerful! But might not be considered “layering”, per se. The things being layered are the elements that shape the sound (another layer of filters, another layer of AM…) not the sounds, themselves.
If what you want is to play two-or-more complete patches in unison, the better solution is just to load each patch up on its own track and trigger them together. The poly config menu makes this easy. Just set the allocation mode to “Unison”, turn down the detune and spread, and check the “use trk sounds” box. Now each track voice will play at the same time with the patch you’ve assigned it.