“Graffiti” as it was called, was indeed a very clever UX solution and it did work well. The concept still exists on modern smartphones for many non-Latin-character languages.
There were Graffiti apps for the Newton as well, but they were not recommended because they wore out one square inch of your screen. That said, I used it on mine because it worked a lot better than the built in recognition.
It’s worth noting that handwriting recognition and writing-based UX is back on the new iPads. They have things like scribbling out text and auto-connecting vector graphics that the Newton did. I hope it works better now. 
The Newton was a classic 80/20 problem. It worked 80% of the time and was magical when it did. However, the remaining 20% is so annoying and inconvenient that it renders the device unusable. Modern equivalents to this are things like autocorrect and smart speaker voice recognition. They are more like 95/5, but still that last 5% is a deal killer for many.
Organic technologies are hard. It’s exponentially difficult to get slightly better so they all plateau at an asymptote just shy of working perfectly.
