Agreed. Most likely coincidence that everything appears to tally. Losses and errors that you are unable to measure. It does appear that your watt-meter maybe doing some calculations based on PF. How well it does, remains to be seen. I expect your losses are accounting for the PF error.
In watt-meters that I have built into equipment, I have started to use an Analog Devices chip that does it all for me via an SPI interface. It can sample away on voltage, current and phase angles hundreds of times a second without affecting the performance of my core processor.
The LEDs will be driven at slightly different currents (and possibly different PWM ratios too) to each other, more to establish the same luminosity at each setting. This will give some gamma correction, but it is more to give equal pure colours to the eye with each of RGB.