Stockfish should do even better at fast time controls, but at 40/40, Stockfish should only be roughly about 20 elo points better.
The bottom line is that 4 games is not enough games to be statistically significant.
If you play many hundreds of games at 40/40 and a large score difference remains, then something is wrong. Make sure that both engines are using the same PC resources (same number of cores/threads, same amount of hashtable size, etc.).
I have held an engine tournament between Stockfish 8 and Komodo 10.1 (both 32-bit) with Arena GUI.
To my utter surprise, Stockfish proved unbeatable, especially in Queen's Gambit Declined. Komodo won not a single game, and drew only when it played white, or if Stockfish used an opening other than QGD.
I wondered if these scores are normal:
Stockfish 8 (32-bit) Komodo 10.1 (32-bit)
Rapid (40/40) 3.5 0.5
Blitz (2/6) 4.5 1.5
Bullet 4.5 1.5
Is something wrong with my version of Komodo? Or is it the usual result, between the engines?
I'd be glad for your opinions.