Here, I have to agree with his opinion that single-player and multiplayer generation are indeed different, unfortunately.
I looked at the new version – it seemed noticeably worse for blue (despite reducing the number of passes). To check, I ran 40 generations, the score was 63:53, and I decided to compare it with the previous one – +20 generations – 33:28 (the first 20 gave 27:28, a 20% difference for 40 generations seems like more than the expected variance – the lunar phase affects the RNG). In total, the old one is 60:56, the new one is 63:53.