I can see it would be useful perhaps to have two shapes then, i.e keep them separate. One for 'safe emissive conditions', and one for typical reflective conditions, assuming D50 white point with optimal lighting, perhaps as you might find in a gallery [and not the average home loungeroom].You can choose other energy rules, and they would probably give other results. For instance, one could assume a reflective environment illuminated with a particular white spectrum (ie. completely flat, or daylight of some kind), and then compute the gamut of every possible spectrum that could be produced by subtracting from the illuminant. But this gamut would exclude large areas that could be produces by quite practical (and safe) emissive devices.
Other people have (perhaps) tackled this quandary. I'm not sure what conclusion they came to.