As it says on our mike position page "Forget all about 1 metre on axis".  When development measure a loudspeaker they have an anechoic chamber in the hope that it will give a repeatable free field responses  The free field response gives you some indication of our how a speaker will perform in different environments; but when comparative testing you're only concerned how alike are the test unit and the reference.  Keeping this in mind the considerations for a production test unit are:

quiet enough to allow the operator to function,

at least as stable as your longest re-calibration period,

and not necessarily approximating anechoic but see the next paragraph.

It helps if the environment doesn't put points of steep slope in the response.  As a slope becomes steeper the change in y increases for the same change in x.  If the environment makes a slope in a unit's response look steeper then differences in that region will be emphasised.