Hi,
I understand your confusion, but MIR "thinks" space differently than a conventional reverb engine :-)
When we record impulse responses for MIR, we don't change their level at all. The impulse (actually: a sine sweep) is always generated with the same volume at the selected stage positions. This means that the volumes of the wet signals are always in perfect relation to what a main microphone in the hall would have picked up in reality.
We cut out the direct signal components of the IRs (i.e., the signal that reaches the microphone directly, without reflections from a wall) when we build a MIR-Venue from it, to avoid phase problems. It is replaced by an Ambisonics-encoded, meticulously positioned version of the dry input signal. _This_ is the part that is scaled in proportion to the distance from the main MIR microphone, but the volume of the signal sent into the convolution engine must of course remain at 100%, otherwise the relation of the wet signals would be destroyed.
... of course, we know that mixing in spot mics is less a matter of science than of art and taste, so MIR allows you to turn off the distance-dependent scaling of the dry gain altogether by disabling the respective feature. I call this the "Pop Mode" (as opposed to the "Classical Mode" in the scenario described above) ;-) Here, the IRs still keep their relation, but the gain of the direct signal doesn't change when you move an Icon on the stage. It's like a very versatile reverb engine, allowing you to pick and place the "voice" of the room without affecting the balance of your mix. Of course, now you have to set the Dry/Wet ratio by hand first, but it stays that way even if you change the Icon's position.
HTH,