Welcome Bob,
thanks for your interest in Vienna MIR Pro.
You are absolutely right: The whole MIR-concept relies completely on Ambisonics, a meta-format which relies on the idea of an coincident microphone array. By definition, there's no runtime delay between the individual channels of an coincident setup (opposed to a spaced multichannel microphone array). - You can think of it as some kind of three-dimensional MS.
I tried to cover the basics in an addendum to MIR Pro's Manual called "Think MIR!". It's available as a download from your MyVSL area: -> http://eu.vsl.co.at/downloader.aspx?FileID=7629 (... please see p.9 ff)
_Any_ signal that runs through MIR is encoded to Ambisonics. It is the only possibility to match the virtual position of the direct signal component _exactly_ with the position if the impulse responses recorded in a hall for the use with MIR Pro, while allowing for seamless interpolation of any in-between positions not covered by an actual recording.
Another great thing about coincident multichannel formats is that they don't introduce any acoustic "smear", thus the positioning that can be achieved is second to none. Consequently any downmix to "narrower" audio formats (e.g. stereo to mono) is free of artifacts which will typically occur with spaced microphone setups (phasing etc.).
And finally, Ambisonics allows for the free modelling of virtual microphone capsules. This is a unique feature and very flexible, as long as you keep in mind that the underlying "real" microphones have been set up as a coincident array. This means: Two (or more) coincident Omni-capsules in a coincident setup will always lead to a mono signal, because there is no runtime delay between them. MIR is able to introduce some "virtual" spacing for the reverb tails (for even more acoustic "enveloping"), but there is no delay between the individual channels in case of the (readily positioned) direct signal and the early reflection part of the IRs.
HTH,