The point beyond which it becomes difficult to change an individual's decoding approach does indeed vary a great deal, and is not going to be the same for a GT individual. A general rule of thumb is that most automatic processes are easier to alter in the pre-pubescent years, but it is not as set in stone as some other cognitive processes, as, if you look at research on adult literacy instruction, it is evident that humans can learn to read effectively well past the point that the brain has matured globally, though with considerably more effort (and some plasticity is still present for quite a long time, as adults who have experienced brain injuries can adapt to some extent).
So I think the bigger issue for altering reading skills is that it is so much easier to revert to one's old, inefficient methods than to practice optimal decoding skills, unless text includes a great deal of novel vocabulary. Hence the value of old o-chem textbooks! It's a question of creating circumstances in which the effort of fighting one's ingrained habits is more worthwhile than battering away with brute-force-memorized-vocabulary, guessing from context, or some other non-orthographic/non-morphological method.
