Of course, the agility here happens by accident, because in the first week of the program increment you realize that something is much harder than anticipated, and hundreds of person-hours of planning are thrown overboard.
Due to badly run prince2/PMI projects people are allergic to anything remotely sounding like a project. So I get the autor when he thinks of ATO == PMO == PMI == waterfall == nightmare. But for larger organisation you will properly need something like a ATO to be successful. Like anything else a badly run, top down driven ATO will fail and make life difficult for anyone involved (this is also made clear in the original McKinsey article).
I like your balanced thinking, in writing the article I was trying to be balanced as well.
Unfortunately when I've seen ATOs in practice it's a band-aid to addressing the actual problems (usually leadership or leadership needing to make structural and large cultural changes).
From my experience it depends on the company culture, leadership and the way agile is implemented and started with in a company.
Big companies with very old and complex architectures coupled with old school separation of business and IT and a bag of bureaucracy. Will properly need a PMO, or SAFe like implementation. Sometimes they are in luck and they are able to start with departments or a business units and move faster. But even if you have that the wider organisation will lag heavily. Most if not all can't become agile in an instance with implementing the perfect agile way of working. Even transforming to an agile organisation is best done in an agile way.