Colors of pure wavelengths are outside of the sRGB gamut and cannot be converted to sRGB, only approximated. Color of infinite temperature, on the other hand, is inside the sRGB gamut so no complication here.
I'm not talking about the out-of-gamut problem. I'm saying that the standard algorithms simply do an awful job, such that you can easily beat them by eye.
This is in the context of trying to produce the closest sRGB colors to a set of LEDs, for use in documentation. No published method could actually do it. I did have to assume the LED was a single-wavelength source, gaussian with specified FWHM, or Lorentzian, rather than measuring the spectrum (I was too lazy to mess with the spectrophotometer), but I don't think that would have made the difference for an LED.
One major issue is that most displays aren’t properly calibrated to sRGB. Once I got two monitors profiled and calibrated properly, suddenly most of these algorithms matched much more closely than I had expected.
This is in the context of trying to produce the closest sRGB colors to a set of LEDs, for use in documentation. No published method could actually do it. I did have to assume the LED was a single-wavelength source, gaussian with specified FWHM, or Lorentzian, rather than measuring the spectrum (I was too lazy to mess with the spectrophotometer), but I don't think that would have made the difference for an LED.