That depends. You didn’t display the duration.
If it’s several uS duration pulses, the falling edge of the first longer pulse is the place to mark time,
then waste half a pulse duration before the for-next loop. If you can roughly synchronise from the first pulse (which is what it’s for),
it takes time for the two clocks to drift apart.

If you were to sample bits as fast as you can, you’d still want to do that from the falling edge of the first pulse.