It wasn’t deliberate. It was more likely scientists being overly cautious about overstating their case. The end results pushed the normal agreed values closer to best case predictions rather than middle.

This was back when the main models needed weeks of super computer time, so only a few were run, and everyone else analysed the shared results.