Luminosity masking is a powerful approach to compositing exposures for sure. One of the major issues with it is the lack of physical reality it often introduces into the tones of the composite image. Usually, the artist will decide that the sky from one exposure will look good with the foreground from another exposure, etc. and mask the one into the other to get pleasing exposures for each element. However, while this may produce technically noise free and pleasing color for each part of the composite, the relative exposures may not be realistically or visually compatible compared to how our visual system would interpret the scene.
This is obviously an artistic choice, but I often look at images that have been produced with luminosity masking and get an impression of the scene that is "not quite right" visually. Again, just an observation.
HDR merging adds light linearly, in a physically correct way, to produce the reference 32bit file. Various tone mappers try to compress the tonal range while preserving the perceptual contrast of our visual system. Sometimes I will tone map an HDR dataset to get an impression of the relative exposures of various areas or features in an HDR scene and then use that result as a guide to a luminosity mask composite, where I can manually do a better job of deghosting or compositing than the more automated HDR merger.
Learning the concepts behind Adams' zone system and trying to establish a visual impression of various zones in a scene will also help in the preparation of the composite image when using luminosity masks. While observing the scene, the photographer can make some notes about their impression of the relative exposure of the sky versus the foreground, etc. and then attempt to preserve this relationship in the luminosity masked composite.