Haha, I can't believe I wrote that. There so much good stuff scattered around here. Lots of different angles to take on neutral, and I think this subjective angle is a good one.
An objective angle is another; but it needs to be supported by the subjective observations. Amps and DAC should always measure flat more or less. If the equipment is not for pro-use (where it would be slotted in long chains of other gear), a little roll-off at the extremes would be OK and probably imperceptible. Distortion and other non-linear characteristics are going to have a larger impact on sound, even in the tonal balance sense, than 1.5 db roll-off at 20kHz. This is what "Hydrogen Audio" guys fail to understand when they nitpick FR graphs from RightMark Audio Analyzer.
For transducers, depending upon the measurement method used, there should always be a target response to achieve neutrality, at least. A flat line isn't necessary that target. At the listening position, the target response I prefer is similar to the old B&K target (
http://www.bksv.com/doc/17-197.pdf) see page 4.
[I've seen HF'ers and a few people here incorrect refer to this as the Olive-Welti curve. The O-W curve is the target / compensation for their headphone measurement system. Sean Olive did not invent this target curve. The B&K target was formally presented to AES in 1974, and sound engineers were already aware and using similar type curve since the 1950s or earlier.] Flat will sound like shit at the listening position. This B&K target also somewhat works with the V2 measurements for headphones on this site.
Now we hear people say speakers should measure flat. Yes, speakers should measure flat if the measurement is taken at 1 meter away with the microphone leveled with the tweeter. That's how Stereophile does it. That's the standard that many speaker builders use. This method is useful because Stereophile or the speaker-builder does not know how your room is going to influence the sound. Hence a "standard". When I build speakers, I always take measurements at the listening position. That is because I can. Studios, when they precision EQ their speakers for "neutrality", also take measurements at the listening position (using B&K target or similar curve.)
Different methods, different techniques, different targets. All attempt to achieve neutrality.
(It's a little bit of a different spin from that found in Jude's sig on HF.)