The first part is dead right...why do you have to do what they do. You can choose your own.
The second part..."mimic a proven design"...What constitutes "proven"? And what came first...the manufactured light of the proof of the concepts in said light?
Did those companies test multiple spectra to then go with the best after rigorous testing to be "proven...NO they didn't. If they say they did...ask for the data of the real test. They release them based on some pseudo science and hunches. Then let their "valued" customers test what they just made and sent to market.
So let's think about what is actually holding you back from making a rough mimic of 30 diodes producing a predominantly R/B spectrum?
30 diodes per cluster at 50w/cluster for chillLED 50w÷30= 1.666w/diode...~555ma/chip.(estimated from the pic you posted above of the cluster)
That could put into circuits a number of ways to get whatever forward voltage you want/need.
And for the final piece of your puzzles to check your work/ideas against what you want it to be...
http://spectra.1023world.net/