freelunch, what the person at Sound Devices told you is basically right--OK, maybe not the part about the tolerance of the resistors so much--but it's true that only some of a microphone's noise output is directly due to its output impedance. A pair of resistors is incomplete as a substitute for a microphone--but using such resistors is still sixteen miles better than leaving the inputs wide open, as some people do.
Just to explain--the capsule of a condenser microphone is a condenser (a/k/a capacitor), like the name says. For small-diaphragm microphones the value of this capacitor typically is between 35 and 50 picofarads. If you place a fixed capacitor where the capsule would ordinarily be, the microphone will no longer respond well to sound or vibration. But its amplifier will still put out the same electrical noise as it does with a normal capsule in place. That's valuable for making several kinds of measurements, since you can inject test signals across the capacitor and thus measure the amplifier's frequency response, distortion, overload point, etc. directly.
Anyway, since the impedance of a capacitor is inversely proportional to frequency, a fixed capacitor has its highest impedance at low frequencies, and therefore contributes the most noise at the lowest frequencies. As you go up in frequency, at some point this "1/f" noise becomes equal to, and then lower than, the noise of the FET or tube or squirrel in a cage--whatever active device is in the first stage of the particular amplifier. From there on up, the noise voltage may vary somewhat as a function of frequency or it may stay more or less flat, but it's the noise of the circuitry rather than that of the capsule's capacitance that you're dealing with.
So the challenge when measuring the noise performance of preamps is to compare their input noise with the microphone's output noise in a way that respects the specific frequency contour of the noise and the huge frequency dependency of our ears' sensitivity to noise. In turn the nature of our ears' frequency dependency is itself dependent on the listening levels (sound pressure levels) involved--so as I said, it's tricky, and any one number can't possibly paint the entire picture.