If I fit only to the data below the big step at 100kbd, I only get another 1mbd of small field capacity:

(Clearly, people like to design projects to produce a nice round-number nameplate capacity).

It's an interesting approach, it reminds me a little bit of the parabolic fractal law. Have you tried to order projects in log(size) versus log(rank)?

Yep. My eye based regression ended up at about 15.5Mbpd or 500kbpd extra. However all those 100 and 200 kbpd items scream marketing numbers rather than engineering.

All of which is unlikely to change the price of fish overall. Even if someone did find a >1mbpd field tomorrow; its still going to be post peak before it came onstream, subject to the same post peak political pressures. Frankly, now, its mass efficiency gains and culture adaptation that has the biggest part to play.