That sounds good, but the devil in the details is how you map the HSV space on to 12 bits.
Presumably they had this as 4+4+4, just as they ended up with for RGB. Now, 4 bits of hue would be pretty limiting, as you'd have just 16 distinct hues compared to the 96 hues that RGB had at maximum saturation and value.
A perceptually better alternative might be 6+3+3, but as HAM works on 4+4+4 this would take multiple pixels to set just H, S or V (and we were hoping to do this with one pixel).
Taking that 6+3+3 idea a step further, the bits could be interleaved so that one HAM pixel gets you to the ballpark, and the following one or two pixels refines it, e.g.:
H5 H4 S2 V2 | H3 H2 S1 V1 | H1 H0 S0 V0
or if that makes the hues too smudgy, maybe:
H5 H4 H3 H2 | H1 H0 S2 V2 | S1 S0 V1 V0
OK I'm living in the past.