The usability of web apps (specifically speed) is improving faster than we can develop our non-trivial apps for this platform. In short: time your version 1.0 with the assumed capability of the host platform at the time of release, plus two years. If you commit to spend a year on AIR development, because ajax or js execution or whatever is not where it should be right now, you're making a mistake.
That's what's bad about pre-obsolescent technologies like AIR - they distract from the future in favor of right now.
The researchers are either incompetent or on someone else's payroll. Either way, they're useless.
Just like an animated human starts to look really creepy as it approaches realism, there's something strange about web-apps that approach the style and complexity of desktop apps but don't quite reach it.
I've seen plenty of great AJAX apps, especially for the enterprise. The problem is NOT with the technology, it's with the programming.
Forrester obviously doesn't understand that. They don't find any good AJAX apps (one wonders where they're looking), so something MUST be wrong with AJAX. Their recommendation: explore other technologies, including Microsoft's.
Run, don't walk, the other way from "research" like this. Look on the bright side - you're not the poor sucker who had to pay for it.
Just a little funny fact: in nearly all enterprise software UIs I've seen, even tab order wan't assigned on dialog panels.
That is not a test for technology, even junk like Oracle survives there.