Except for the RX-V377, above graphs are from the S&V top pick list linked below.
http://www.soundandvision.com/content/top-picks-av-receivers#ZuV2dJsZZHQ7kTvG.97
If you study them carefully, their shapes/slopes and the THD+N from 0 to 10W, you will see why we can't just say going from a 70W AVR to a 140W AVR the difference is just 3 dB. Now if we all believe anything below 1% THD+N is not audible then the blanket statement could still be true, but keep in mind the above graphs are for an 8 ohm resistive load only. By extrapolation, one might reasonably assume the flag ship or near flagship models such as the Denon 4520, Marantz 7008 Yamaha 2040, NAD787, Pioneer SC89, Onkyo 3010 as shown above will likely have way better THD+N figures and shapes than the entry level models.
The data and graphs from AH right here are way better and with more details but there are as many available, so I am forced to go with S&V's (formerly HTM).
My point is, up to the point of diminishing return, we probably should compare power outputs and distortions together, even in pure direct mode (that's a given). The 70W AVRs tend to be the entry level models, as they drop the power rating, their other vital specs, such as THD typically drop as well. Some young and well trained ears such as Seth's could quite possibly hear the resulting differences.
All comments are welcome, hopefully no sarcasm and we can always agree to disagree.
And yes, I know THD+N is just one of the vital stats, but that's all I can easily find for now; and it certainly is not one that we should ignore. I do believe it is a major one to look at, as it is included in pretty much all bench tests.
Found this site that has some good links:
http://www.parallelhomeaudio.net/TypesAudioDistortion.html
They talk about a type F distortion, or Mental Distortion, it is hilarious.
Another interesting one by Roger Russell:
http://www.roger-russell.com/truth/truth.htm