Exactly, the lower limit, when hydrogen has escaped into the air, not when air has leaked into the contained hydrogen. The article makes clear that above 75% H2, ignition is impossible (but only if you read it).

"Traveling along grain boundaries" can only leak trace amounts. With, as I said, positive airflow, trace leaks can be constantly diluted below the concentration where ignition is possible.