Axis dynos are pretty good at simulating real world conditions if you set them up correctly, though I'm always a fan of making corrections based on data in the "real world" after the fact, if necessary. They generate wind speed directly proportional to wheel speed and throw it at the front of the bike. I actually get ram air compensation up north of 80mph, and it simulates drag (almost too aggressively on my set-up!). The wind is aggressive enough that I have to hide behind the screen on big pulls.
This customer will likely show up with a good, unrestricted base map. I'll smooth everything out unified to get a good baseline, play with injector balance, then move timing around to see what it likes. Once I have an established good baseline, I'll see what autotune does to the balance un-unified (provided the customer doesn't mid), compare what it does to AFR/power, and correct afr if necessary. The dyno will repeat well under 1% so long as run conditions are held even, so I'll know exactly what sort of power differences are occurring, and where.
If I'm using woolich stuff and the customer has the AT hardware, I'll usually make successive pulls in various conditions to let it get about 95-97% of the work done. After that, I'm looking at the actual torque output vs specific changes in timing, AFR, and injector balance... and usually playing with different ETV mapping for different modes to allow for different power delivery strategies. 99% of the folks around here are primarily straight lining, and there is a lot of advantage to be had in a bike that drives hard off the midrange but doesn't wheelie excessively. We are at a fairly high altitude (4300ft), so we can be more aggressive here than most... but traveling down to low elevation events requires a different mode/approach. It's always a balancing act at these power levels, until you add a little swingarm. As far as AT goes, it's not bad with the woolich stuff, and it does get it 97%+ there every time... but getting that last little bit requires precision measurement under controlled conditions. Most AT is rather terrible at honing in on that last little bit, and without measurement devices you can't tell anyway. Trying to feel a 2hp difference at 200hp peak is extremely difficult for us mortals. Trying to observe a change in trap speed can also be difficult, as real world conditions constantly change (rider errors, road conditions, weather, etc). A little bit of temperature or wind skews it all to hell, for better or worse. There is more MPH in the half mile to be had by good body position than there is in injector balance, for most people.
And to add one more thing to the mix: what "feels" fast isn't always necessarily the case. I can put a big hole in the powerband right before peak torque, make the same peak HP, and the bike feels like some rowdy firebreathing animal... where smooth torque over the same area will actually feel slower, but yield consistently better trap MPH.
My prediction is that it won't do much. After all the work to get the baseline 98-99% of the power will be there (save what you could get from actual, individual cylinder tuning... which no one wants to sensor up/pay for in dyno time). I'm not saying there isn't power to be had with injector balance, not at all, just that letting autotune have at it after I've already tweaked it will likely yield diminishing returns. If improvements do show themselves at that point, it will be in the 1-2% magnitude, and almost exclusively after 8K rpm... which comes out to 2-4hp at peak. Once an AFR is correct, there is very little change in power output within a relatively large range around said AFR. If a motor makes best power at, let's say, 12.9:1 (.88 lambda), it will stay within 99% of that power within a half AFR point (+/- .03 lambda). Now, there are some things going on here with charge cooling and charge air density/composition that realize some gains, but that has less to do with AFR and more to do with WHERE and how much fuel is being introduced relative to its position from the intake valve. This is a dark, deep rabbit hole my friend.