There are many possible explanations for the discrepancy. The failure might have been in predicting the performance of that base model 911. Maybe the data they used was someone who was very slightly less urgent on the manual transmission.
The failure might have been a failure to predict the performance of their own product when it was still in the late stages of tuning. A last minute change to tyre compounds or stability control algorithms could have made enough of a difference in this very very close race.
It could be argued that in a fair comparison between cars, the stopwatch should begin when the driver slams the pedal, not when the car itself starts moving. The latter is how races are timed, but I'd argue that the former is more honest. Including the split-second delay of the ICE engine might make enough of a difference.
My zero-evidence guess is that the Cybertruck team predicted the outcome in simulation, hyped everyone up about this as a visual demonstration, then spent tens of millions of dollars replicating the outcome in the real world, buying two Porsche 911s, renting a film crew etc. Except they couldn't replicate it in the real world. So instead of admitting the wasted money/time, they covered their asses and misled their boss.