Unfortunately, spammers do have way more advanced technologies to try and guess working addresses. Libraries like this generate mostly very improbable addresses: you would get way better results just by taking lists of most popular first names and family names and concatenating them with some well-known web mail providers.
The bigger problem, in my experience, is when you accidentally spam people yourself, while testing some function that somehow generates an outgoing email. I think almost everybody that's ever worked on code that sends emails has - at one time or another - ran a job that somehow wound up sending 6,712,374 "test" messages to every user in the production database. :-) Dummy addresses are better, but it's better still if they use @example.com so you know there's no way they'll actually get delivered to anybody.
I'm gently disturbed that it generates valid email addresses. You pick a random name, follow it with an @, then add a random word, and then add a TLD. Obviously you're not responsible for how people use the tool.
Best practice would be to use the domain example.com