That is the problem with LED, it is a different technology than an incandescent filament.
Whereas dimming a filament causes a orange-shift and decrease in color temperature, ordinary dimming of an LED can result in a slight (but noticeable) color shift, as the blue wavelengths from the diode become a little (2-4 nm) longer.
...for some values of 'noticeable'.
Another little technical problem is that it's challenging to design a circuit to drive the LED that is both capable of ordinary dimming and is efficient (although I am sure it is no doubt possible).
It's debatable how significant ultra-high efficiency would be when running at significantly reduced output.
If a driver was 90% efficient at max power but 50% efficient at 5% power, the actual energy lost at 5% output would be small, and in a situation where that was significant/costly enough to care about, some solution like having multiple drivers would be possible.
PWD is a strategy that seeks to avoid these problems by simply modulating the LED on and off at high frequency, but this leads to the problem of flicker, which, according to some reports that have researched into this, can really have insidious effects on someone in a working environment even if they are not consciously aware of it. Eye strain, difficulty concentrating, agitation and such.
PWD is the wrong direction to be going in, in my opinion, but engineers have never been known to be considerate about the ergonomics of the things they design.
Flicker is significantly an issue of the frequency chosen. With a high enough frequency, it may only be potential stroboscopic effects on moving objects which are significant.
As for engineers, much comes down to the design brief, which should be written with an eye to expected customers - what they want, what they
think they want, what they will like or dislike in practice, and what they are prepared to pay.
There are clearly some people who would complain whatever was done, and they may well be the last people anyone is likely to care about, or even desire to have as a customer.
Another obvious strategy would be to just turn off some of the emitters.
That does depends on the light fitting and optical arrangement and the range of desired outputs.
If there are multiple emitters lighting up the same area, for outputs of 'at least one emitter', it is possible to have some emitters on and some running PWM, meaning less intensity change between PWM on and off states, which can significantly lessen stroboscopic effects on moving objects.
With an eye to flashlights, even when there is a single emitter, where there are parallel constant current drivers (such as the trusty AMC7135), it is possible to do a similar thing there - power control by turning one or more drivers fully on, and having one driver running PWM to make up the fractional part of the power output. I've done that, and it seems to work quite nicely.
