In the Mini review, SPCR found that a ~19dBA fan (Slipstream @ ~800rpm) was too loud while running in a ~23dBA system in a 19dBA environment. Reducing the fan's noise by 1-2dBA dropped its noise level below another component's, giving a significant improvement in overall noise. But SPCR still managed to measure a .5dBA increase in the system's sound output when the fan was turned on at the slower speed.
I think your mistake is to rely just on the numbers. That's why SPCR supplies audio recordings of the fans they test.
A lot depends on the character of the noise. As I have already said, and as many others have found, I do not need to run my fans at 5V for them to be inaudible. I can't hear them at 7-8V, even 9-10V in a P182 with some Yate Loons. I had my hearing tested recently and it's above average for a 28 year old. I live in a fairly quiet house.
This is the point you seem to be missing: it doesn't matter what differences you can measure in a soundproof room with sensitive equipment. What matters is that I can't hear it in my bedroom when sat in front of my PC. If I can't hear the fan at 7V, there is no point running it at 5V or any lower because it will make no audible difference to me but will be pushing less air.
I'm not sure how else I can state this argument to make it any clearer. In fact, you yourself seem to be supporting my argument with the statement "The conclusions to reach are that implementation counts & that you cannot judge an entire system solely by its components." The only difference is you seem to have come to a somewhat illogical conclusion for some reason.