I’m just on chapter 21 of LL pars I and I’m enjoying it quite a lot but I can see that from the point of view of modern language acquisition theory the series should be three times longer, that is the new grammar and vocab should be spread of the 3 times as much text. You can climb just as high when the slope is long and gradual as you can when it’s shorter and steep.
My fantasy is a series of graded latin novels modeled on the Cambridge University Press excellent series of graded English readers. They would start at a level about 3/4th of the way through Familia Romana and take through the grammar and vocab of LLPSI part 2 but spread out over much more text (I’m thinking at least a half million words total if not a full million), all of it entertaining fiction. They would include murder mysteries set in ancient Rome (like the Gordianus the Finder and M. Didius Falco books), stories built around the exploits of a Roman soldiers, etc. All in increasingly more and more advanced latin providing massive amounts of comprehensible input. They wouldn’t need to be “per se illustrata” a la Orberg. A glossary in the back would be fine, but you would need enough levels (maybe 6 or 7) to make the transitions from one to the next not to difficult. The goal would be to make the reading experience all the way through as smooth as reading one of the middle chapters in LL pars I right before the slope starts to increase.
The theory is right and it would work but of course will never happen, not enough of a market for it, I suppose.