For speaker response, flat, that is 1:1 with the input, sure. However, when actually listening to music, people like non-linear response curves beyond what averaging and smoothing many rooms very jagged individual responses predict. In Toole's book, Fig 12.7 (pp 351 in my edition) shows the "Subjectively-preferred steady-state room curves" from Olive, and none of those are straight lines, much less flat. For trained listeners, above 500Hz or so, it's pretty linear. Below that though, it has a pronounced hump, and for untrained listeners, which includes almost everyone, the hump is very pronounced plus there's significant treble boost. I'd expect that in real rooms, people's tendency to compensate for room acoustics would keep this non-linear preference. So, for systems set up for a pleasing listening experience for most people, I'd expect the house curve, room curve, balance, whatever you want to call it, won't be linear, and certainly not flat. It also is likely somewhat idiosyncratic, perhaps evolving over time and experience.
My point isn't to be pedantic about this. IMHO, with the widespread advent of pretty much audibly perfect DSP corrected speakers and electronic components, the emphasis in home HiFi testing is going to shift more in the direction of "presence". That is, the overall system's multichannel effects that make recorded sound more like hearing live music in a good venue. My suspicion is that the "presence" effect is why Olive's curves, tested in mono, aren't linear.
Here's a presentation by Sean Olive, et all, discussing room curves for trained listeners:
https://drive.google.com/file/d/0B9...ew?hl=en&resourcekey=0-uSgBBin0zcCtmpgvEsdskQ