The integer encoding of Lab by means of three Cartesian coordinates (cuboid space), as given in Photoshop, bears a low Coding efficiency (just 35.1%). Means there are many many combinations of L, a and b which appear to be outside of the blob like Lab space describing the gamut of human vision. This is illustrated in full 3D glory on Bruce Lindbloom’s website. The point is that we still can make use of these "not real colors" as Bruce likes to call it, or "impossible colors", based on the given mapping functions applied under the hood, thus turning everything to real-world colors.
I would agree we can define 16.7 million colors definitions using 8-bit per colors but not necessarily 16.7 million perceivable colors. Many of these values are totally redundant. Can you see the difference between 0/0/0 and 0/0/1 on a display?
On the other hand, color is a perceptual property. So if you can't see it it's not a color. We define colors based on perceptual experiments. Color is not a particular wavelength of light. It is a cognitive perception that is the end result of the excitation of photoreceptors followed by retinal processing and ending in the visual cortex.
A coordinate in a "color space" outside the spectrum locus is not a color. Some refer to these as "imaginary colors" (Dan of course had to make up his own term) but this is by and large also erroneous (you can't map an imaginary color from one color space to another as the math (and experimental data) for each color space breaks down outside the spectrum locus.
My beef is the made up terms or modified terms that only Dan uses when existing terminology has been in place. The ICC doesn't have any such definition for "False Profile" and I don't know anyone, other than Dan who uses the term "Ultra Wide Gamut Spaces" to describe say ProPhoto RGB but somehow, Adobe RGB (1998) slips under that made up heading. Or "Range-opening routines" (another made up term). Everyone else would call it tone correction.