In my experience an average speaker well positioned in a good room, will always measure better than a great speaker badly positioned in an bad room. It is my experience that we listen more to the room than to the speakers.
So, meaning all respect when I say this, the above isn't supported by research. I say this as a reviewer who mostly listens but also measures.
Expanding on what Amir said, if you read Dr Toole's
book, which is mainly an overview of much of the existing research, you'd see several important things pointed out that make obvious why anechoic measurements are so important:
1) IMO the most important one: Humans have a remarkable ability to hear a speaker "through" the room. The room colors the sound, but the science suggests that above the Schroeder frequency (basically above 300-500hz ish) in most rooms the primary things we hear is determined by the anechoic response.
This seems related to how people even understand sounds in reverberant places in the first place. Imagine listening to someone's voice in different rooms -- you can still tell its the same voice with all its qualities and quirks even though it would 'measure' different in different rooms and technically sounds different.
The direct sound remains dominant in most home-sized rooms too. Humans are able to separate direct sounds from room reflections very well, and when you enter a room, you are able to quickly assess it's acoustic properties and use this to separate the two.
In one experiment, there's even evidence to show that this is a process that takes a some time for adaptation, and that the first few moments when changing acoustic environments rapidly, you're likely to have less clarity separating these sounds.
2) Below the schroeder frequency, the room have an increasing importance, which is likely why positioning and such does have such an impact, because the bass has an effect on how we perceive the rest of the frequencies. Nonetheless...
3) ...when it comes to
evaluating speakers, measurements are still useful because there is significant evidence to show that even a speaker will sound and measure different in different rooms, the preference
ranking for the speakers will tend to remain the same. In other words, it's true that some rooms and positions are better than others, but theyre likely to affect most speakers similarly.
4) Two ears and a brain are not a microphone and a computer, and what we see in a graph does not fully correlate with how we hear things. Research suggests that the measurements that most correlate with how we hear things are anechoic measurements. Experiments where researchers try to assess preference based off of in-room performance are less reliable than ones based on anechoic data -- even though in both cases the listeners are listening in a normal room. As you factor in the room, the data gets worse at predicting what people will actually like.
5) The above is largely because in room measurements don't take into account how the brain perceived data coming from different directions and with different timings differently. For example, two speakers can be calibrated to have an identical in-room response at the listening position, and yet still sound quite different largely because of their directivity properties. But it's also because in room measurements will tend to lack the resolution to identify narrow resonances and the like.
All this is to say that even though a speaker may be affected by the room, all the science shows that above bass frequencies, it is the speaker itself that seems to dominate our perception of tonality
and our preference for speakers. There are some blurry lines -- I think wide directivity might sound better in some spaces tthan others, for instance -- but by an large the above explains why measurements are so important for assessing speaker performance.