Strictly speaking, it wasn't P. being expanded to PRINT at the time of LIST but the byte representing the tokenised PRINT being expanded then. The P. was turned into that byte earlier.