You have got to be kidding me. What does DIY have to do with anything? Only integrating spheres used by DIY people count?
It's "they" as in the people who manufacture these leds and create the datasheets.
BOBBY_G always says that these are just indicative numbers. It's cool that he does this and it gives us extra insight in how these COBs compare or how they would perform in real life situations, but by no means would those replace a datasheet or actual PPF measurements.
That you don't understand the consequences of all the possible grounds for errors I listed does not mean they don't exist. The binning alone can already cause big errors. These COBs are not created in discrete models of a certain efficiency. "They" produce them and then measure in which range the efficiency falls. So a CXB3590 BB would produce a minimum of 10000lm, but in reality that can be anywhere between 10000lm and 11000lm. That's a 10% spread. What if the COB tested is on either side of those extremes? It gets worse if the COB you compare it to is at exactly the other end of it's bin. You might conclude there is no difference between a CXB3590 CB and CD bin. Or vice versa that there is a 20% difference between the two.
You will never know if it should be 0% or 20% (or anything in between) unless you test a lot of them from different batches. Which obviously is impractical if not impossible.
Or just use the datasheet and only use those measurements as an indication to check if the datasheet would be sort of accurate within a 10% to 20% margin.
What BOBBY_G measurements show is how the COB deals with heat management.
