It's the same concept as adding artificial tape hiss to a recording from which it has been digitally removed: you get back a sense of 'analog-ness', and it somehow 'feels' like there's more detail in the high frequencies. If you manage to dial it in right, the constant low level noise throughout the recording helps glue the whole thing together sonically, and it no longer feels wrong as it did when the hiss was just missing altogether. It doesn't actually add any detail back that was removed by the noise reduction, of course, but it can help make a better listening experience.

Tape also adds slight compression and harmonic distortion, which also help give that sense of rightness even while technically being less 'accurate'. It's an odd phenomenon, that our brains seem to prefer the slight imperfections of analog, both for the visual and aural realms, because they feel more pleasant and 'real' than the cold over-precision of digital. But since analog has so many physical limitations, while digital is much easier to work with and is limited only by the accuracy of the equations programmed into it, a blend of techniques from both domains is required to achieve the best possible results.