Point 3 - I didn't know this.
Point 4 - What is BOM?
Point 5 - Like in a phosphor-based LED chip, couldn't a maker miniaturize the discrete R-G-B diodes on one chip?
4) "Bill Of Materials" - all components needded to assemble the thing and to make it working. In other words it is complex.
The RGB concept works only when you want to display majority of all colors (you can not display all in the "horseshoe" but only a triangular part of it, but that contains the ones mostly present in scenes). So the only thing you need is they excite the eye sensing cells the same way the real scenes do.
But for an illumination you need something else: You need to make sure the light reflected off objects illuminated by that light source excites the eye the same way as when the object is illuminated by a daylight.
It may seem there is no difference at first, but if you dig into it, you see the difference is VERY significant:
Assume you have monochromatic RGB LEDs, so each one emits one wavelength, so it excites its corresponding eye cell type. So by varying the intensity you may go across most of the visible colors (most, because the spectral sensitivity of the eye cells is pretty wide and overlap a lot, but that is a detail now).
But if you shine on an object reflecting just one narrow line between those LEDs (e.g. yellow), this object will be yellow on a broadband sunlight, becauise the sunlight spectrum does contain emission matching the reflected one, so that part reflects off. But this hypothetical object will look completely black under that RGB, because the LEDs radiate at red, green and blue lines, but not at the yellow one.
This example is a hypothetical extreme, in reality the reflectivity of real objects is more broadband, but having just 3 lines is too few so the color gets distorted.
Normal CRI80 rare earth (tend to generate narrow spectrum lines) fluorescent phosphors have to radiate on 5 wavelengths to work well with most real life objects.
LEDs use wide spectrum phosphors covering red to green, the blue region uses to have just the primary blue LED line so has gaps there. But nature does not offer much narrow reflected spectrum objects, so for normal use it is OK.
But there is another problem with RGB:
The LEDs tend to loose efficacy as their temperature rises, but each color does that with different slope.
So when blue degrades with 10..20% when warming up to ~60degC, the red tends to "fall down" to 1/4. That means the ratios among the color components changes as the LEDs are warming up. This effect needs to be compensated by the driver to get reasonably stable colors (most applications don't care, if all that is needed is changing colors), which need good LED temperature sensing and good performance charts of the used LEDs. And the accuracy of this data is usually the main cause of the off-color output of the RGB sets.
And for the multiple colors on one chip: This is impossible because of the laws of physic. The thing is, the color depends on the electronic band gap of the semiconductor material. So for any given wavelength you need dedicated material. But if you want a single chip, you need to select one material for that chip.
The manufacture makes many inch diameter wafer (old used 4", modern are up to 20") of that material first, the structures of many 1000's (typical 1W LED is 0.5x0.5mm, so we are talking about 70k 1W LED dies on a single 6" wafer; a manufacturing batch tend to contain 24 wafers processed at once) dies are made on it and only at the end it is then sawn to the individual chips.
So most process steps, which are expensive, make 100000's chips at once, so even when the process is very expensive ($12k to process the batch, so $500 per wafer), this cost gets diluted among all the product units, so those units then seem to be dirt cheap (0.7 $-cents per a single LED).
