I would not trust Sofirn to be very accurate either, but the difference between smooth and OP reflector has been measured before to be limited in specific circumstances: I would expect the difference to be smallest with large die leds in small reflectors and largest with small die leds in big reflectors. Back in 2015 I made a rig for reflector testing and although I never did many actual measurements (got bored, I found making the rig more fun than actually using it) it may be interesting to check it: Ongoing project: reflector testing