22 cheats out of 108 is big - and the real proportion may have been even higher, given that it seemed like he caught only either blatant cheats or conscience-stricken/less brazen cheats (the latter category from when he acted people to own up). And it's a lot of work, and I've been around this process at a couple different institutions and seen how tough it is to see it through to an appropriate conclusion.
However, all this wasted time, and all this aggravation, wouldn't be necessary if all the other professors were doing it too. The only reason they're at 28/108 or higher is that people have obviously been getting away with almost anything.
The author has buckled (understandably) as being the only hard-ass in a environment where everyone else is getting away with it isn't feasible. From his perspective, I can see why he didn't fight this, but if he'd stuck it out for another year or two - and showed the next round of incoming classes just how ugly it will get - he'd have gone back to 'normal' cheating rates rather than 28/108.
The idea that you can always produce assignments that are 'unique flowers' that can't ever be duplicated year by year by cheating students has its own pitfalls. One problem is exactly that these assignments _are_ unique flowers, and might turn out to be systematically too hard or too easy or too vague. An advantage of the rather mechanical, "near-clone of last-years stuff" approach is that you can learn from last year's assignments and tune them onto the target. Faculty shouldn't be dumping the same material without fixes every year, but in many fields it's insane to expect that they have to prepare all new assessment material every year. They just don't get paid enough or get enough time for that.