> But most colours can be associated with a primary wavelength… except purple. So by that definition, they don’t really exist.

And white, and black. Physically, you'll always have a measurable spectrum of intensities, and some such spectra are typically perceived as "purple". There's no need to pretend that light can only exist in "primary wavelengths".

Even if there's no empirical way to extract some 'absolute' mental notion of perceived color, we can get a pretty solid notion of perceived differences in color, from which we can map out models of consensus color perception.