I've found quickbeam's setup to be accurate within 5-7%, certainly good enough for me. To get that level of accuracy you have to be very consistent with the setup and use several lights of known lumen output to arrive at a meaningful conversion factor for a particular setup. And it's only useful for white lights. Colored LED and such aren't measured accurately by common light meters, as the CPF benchmarking lights have shown.
I'm not a big fan of ceiling bounces as a way to estimate lumen output. While it appears to be a good method, and it may very well be for some folks, I've found that it is prone to unintentional subjective bias. There's been several times where I thought 2 lights related to each other in a certain way, only to have my light box/meter prove it to be otherwise.
Not many folks using quickbeam's approach...which is too bad, IMO.