I thought I would weigh in on this topic. Lets start with the CXB3590 by Cree, COBs certainly have their place in lighting and agricultural lighting but for specific uses. First lets look at the lm/w, which I agree are not the correct units of measure for plant growth, never trust Cree's advertising you have to look at the datasheet! The CXB at 3000K (spectrum speaking a good starting point for plant growth) driven at 36V 2400mA yields 86.4W and hot-binned (once the LED is warmed to 85C case temp) yields 11,000 lumens which is 127lm/W not 300 lm/w. This 127lm/W assumes one can keep the junctions of the COB at 85C which for such a small source is no simple task. Keeping this device cool requires a low thermal resistance path to ambient environment. The CXB is far from 66% efficient it is closer to 35%; given this efficiency 35% of the energy is leaving as photons and the 65% is converted to heat (86.4W)(.65)=56.2W of heat. If you start with an ambient temperature of 25C in order to keep the LED temperature at 85C you would need a thermal path or cooling system capable of 1C/W, which is not impossible, but no trite matter. Another design consideration for the COB is achieving a homogenous light pattern at the plant canopy. Light which does not strike the plant is a waste regardless of the uMol/W of the light fixture, and a single source COB with such a large LES is hard to shape in a homogenous manner, and very hard to change as your canopy changes throughout time. Albeit, designing a fixture with multiple LEDs is far more challenging then the COB path, but optically and thermally speaking there are some major advantages to this topology.
The last point I would like to touch on is a word of warning of trusting the quantum sensors mentioned in the previous post. Quantum sensors are an incredibly valuable asset to the would be horticulturist, but beware of the readings when not using sunlight. The sensors are designed for specific sources they are measuring, if you look at the sensors documentation they give different weighting depending on if the source is HPS, fluorescent, or Sunlight. The best tool for measuring the LEDs par value or PPFD would be a spectrometer. The spectrometer also gives the added benefit of being able to see the exact (or as close as you can with measurement) to which nm of light are being emitted.