I've personnally experienced rolling dyno runs in 4th gear also. But 3rd seems to be the norm, which might actually lead to some issues : some cars have torque truncation in 1st, 2nd and even 3rd gear due to gearbox protection. I don't know of any US-sold car like this, but the EU Ford Puma 1,7 had torque truncation up to 3rd gear because of its weak Fiesta based gearbox. So if you dyno it in 3rd, you won't measure full torque.
And it's never done in first gear because you would be in fluctuating transient conditions all along the torque curve. 3rd or 4th give you conditions a bit closer to steady state.
Imagine how much better rotaries would be today if all car manufacturers had been developping them competitively for the last 30 years !