Mark,
It is difficult to refute the actual test results of you and Andy... and I am still thinking that perhaps there is another explanation. (In fact, I have a theory I will share in a moment.) At most, there is a possibility that the MAP values can have a very small effect on AFR... but not nearly to the degree that you suggest.
Once again, the NSX like virtually every other fuel injected car out there will be running in "closed loop" the majority of the time. In closed loop, there is no need for the ECU to even poll the MAP sensor - although perhaps it is taking a look just for collaboration. In closed loop, the ECU has pre-programmed tables for injector pulse width and it is using these and then verifying the results by looking at the O2 sensors. If it has the pulse width too long (rich) or too short (lean), it gets this feedback via the O2 and it can dynamically make the adjustments needed. This is why the system is called closed-loop. It is completely self-reliant... applying values and able to get and learn from the feedback. Think of it as a two-way communication.
Furthermore, in the event that the ECU cannnot, for some reason, achieve stoichimetric, it will issue a fault code and CEL. There is absolutely no ability for the ECU run leaner or richer in this closed loop mode. The stock O2 sensors are essentially binary devices arouund 14.7:1. In theory, the ECU can be either faster or slower in the way that it applies correction to errant AFR values and that is why I said that very small differences could potentially be possible. (Personally, I have my doubts about this too.)
To further make my point, the whole reason that we have this second mode, open-loop, is to allow the motor to run under richer conditions while under load. If it were possible to achieve various AFR's in closed loop, there would be no need for open loop but the fact is that open-loop exists... for this specific reason. In open-loop, the ECU's desire is to run at a richer condition, say, 13:1. Since there is no way to achieve this by monitoring O2's (only valid around 14.7:1), the car instead must rely on a variety of other sensors. In open-loop, the ECU is polling these critical sensors, MAP, temperature, etc and then it is using a look-up table to determine the right pulse width for the injectors. If the ECU is wrong, there is no way for him to know. Think of this as one-way communication. OK, so the best that can be done in open-loop mode is to do the best possible job to insure that the sensors inputs are valid. If the sesnor inputs are valid, then the ECU assumes all is well.
Several of the SS experimentors here have probably at some time put a value of greater that 13.8 into one of their MAP cells. In fact, the early tables from MB himself had a few of these values populated about. The trouble is that that the ECU NEVER expects to see a MAP value greated than 3.5v. If it does, it assumes that the MAP sensor has freaked and is therefore not reliable. The ECU will immediately throw a fault code, trigger a CEL, and go immediately to limp mode. (SS map values are simply MAP voltages x 4. A cell value of 14 = 3.5v.)
In summary, there is nothing to support the theory that MAP values can/will change AFR's during closed loop. In fact, I believe that neither you or Andy have any direct evidence of AF's being effected... but you do have the data point that changing the MAP values has enabled you to pass smog.
OK, so here's my theory. I claim that you are not effecting AFRs (for all of the aformentioned reasons) but you may be effecting the EGR valve position which will definitely change your emissions. In the NSX (which I believe is somewhat unique in this regard), the ECU computes (from external sensors) and then positions the EGR valve position. Unfortunately, the ECU also checks to see if the EGR valve has moved to the position it instructed... so it is hard to fool it. So, my theory is that in changing the MAP values via the SS box, you have both caused the EGR valve to go into a different position that is much more smog compliant.
On a final note, I am installing my FJO AFR gauge this morning. This will allow me to have real-time AFR feedback at all times. I will definitely be experimenting with SS map values and I will report back my findings.
Andy, nsxnut, please don't view my input as argumentative. I'm just trying to debate with you for the benefit of all. Other views/experiences are welcomed and appreciated.