There is a huge selection bias in these groups. 3.5 years without graduating is likely to be students who are doing poorly since if you were a top tier student you aren't going to quit 0.5 years off from a degree under most circumstances. People quitting university after 3.5 years most often have a number of failed courses and aren't 0.5 years from a degree.
Another example then: college students forget a lot they learn inside of a few years. The only thing they hold for a significant amount of time is the prerequisite knowledge they needed to learn everything else they were taught. This applies especially to subjects largely unrelated to their degree / interests.
I think are two different stages we're conflating here.
Education, in the resume / HR sense is absolutely a signal.
Education, in maintaining and excelling in your profession is knowledge.
It's lamentable that degrees are used as filters, but I wouldn't trust someone to do an electrical engineering job who didn't have a related degree. There's simply too much to learn. (Yes, exceptions do exist, but they're not common)
I'd argue the more pointed questions to colleges should be: if you're not equipping your graduates with the latter knowledge, then what are they buying? And why should they continue to pay you?
Doesn't that book have a whole chapter espousing hyper-libertarian points about why child labour is good?
For instance the difference between 3.5 years without graduating and 4 years + graduation is immense, whereas in a non-signaling model, they should have very similar outcomes.