“Subtractive distortion”, never heard that before.
I can think of a similitude.
Let me play devil's advocate for a minute...
Say I have a camera with a 48Mpixels sensor.
I can either record the image as-is, in which case, I have a crazy resolution, but quite a lot of noise, or, say, down-sample it to 12Mpixels by averaging 2x2 pixels, and decrease the noise.
Of course, I then loose some details. How much of those details are relevant is dependent of what you want in your image. Sometimes, details are what matters. Sometimes, a smoother, more immediately readable image is what you want (you may then go even further in the averaging process).
You may also want to add some "high frequency noise" to the image to make it look better, in case you have recurrent low frequency "distortion" pattern in your image (think moire, as an example) and you want to attenuate it.
Measuring dynamic range, resolution and so on, you would obviously see differences... that may not be as relevant as the "effect" on your output. You may well like the "broken" image better.
I guess same rules may apply in audio ?
Of course, just like in audio, I'd rather have the pure 48Mp image and decide by myself what I'll do with it in post treatment.
Salt to taste.