As you point out in your blog, even if the numbers are true the normalisation process removes almost all the interesting features.
Is the dip evidence of a reducing cohort half life? Young teachers leave after 5-10 years and are replenished by return to work or career change teachers much later than in the past. Or the younger teachers now have a shorter half life so the gap between them and the second cohort is exposed. Very tricky to unpick what is going on.