>Similarly, orange may well be red + yellow in scientific
>or artistic terms, but no natlang is going to see it that way.

Why are you so sure about that? I think it's like stating
that no natlang is going to see purple as red + blue (German
does: blaurot). If you lack a specific term for orange,
referring to that color using its basic visual components,
red and yellow, seems to me only natural. Chaucer referred
twice to orange as the color "betwixt yellow and red". And
even nowadays having "orange" in English, "redyellow" has
also been used for example for poetic effect - just search
the web:
"toward the far shore, the dustgrey sky
above the first redyellow fires
scattered on the green October hills"
"Sunset over Cuba. A universe of stars in the darkness
below. Crescent moon hangs in deep blue sky. A bright
redyellow horizon splits my vision."

>> Hmm, 'gold' is really a surface property affecting the kind
>> of reflection but not the wavelength, so that is no color.

It is a concept that combines color with visual texture.
In fact, in real life it is often difficult to dissociate
both aspects, because we perceive textures visually by
means of color variations within the field of the surface.
"Gold" is a color-texture combining an overall yellow basis
with shininess, which causes that yellow to fluctuate in
lightness widely from almost white to almost black.