The problem is in the complex calculus derivative curve programming that is required to program the IRIS's interaction with the gamma and average APL. This type of programming isn't for the faint hearted, and I guess they keep hiring the wrong people with the wrong skillsets to program the IRIS's.
Somehow Sony found the right person or people to do the job. It appears most of the IRIS programmers are just randomly throwing generic algebraic equations at the problem and not realizing there is actually a hard formula to use (hmm). Though the base formula probably needs some interactive tweaks and adjustments on the way, there is no excuse, reason, or rhyme for an IRIS to be as poor as some of these IRIS's are. Maybe they are outsourcing the IRIS programming for $5 hour. They need to hire someone that worked on Paint Program software for years (like Adobe Photoshop) or a game developer that worked on a low-level game engine, or even a pure math programmer could give it a shot. Personally, I would hire all 3 and have them all write IRIS code, then see how comes up with the best, and send them back to the drawing board to do a second version. Too bad they keep hiring rookie programmers to try to program the IRIS's. The other frustrating thing is most IRIS's only have 1-2 modes you can choose from, some maybe 3-4. Personally, every IRIS should expose about 5-6 variables that you can adjust as modifiers to the formula, then people could tweak as they prefer.
I personally would not use a Dynamic IRIS as my decision making mechanism, as they generally don't help the image enough IMO to put any bets on it. To me even more important is a manual aperture to be able to adjust the brightness in steps so that you do not have to fiddle with attaching / detaching ND filters.
The magic of the IRIS is most noticed on plain black nothingness or maybe just above that, that's the downer to it. Many IRIS's don't kick in correctly even on credits.