Paul and others will know better, but my guess is that the rounding is tube "cut-off" -- i.e., the tubes stopped conducting a bit prematurely (since tubes invert signal, the plate voltage is highest when the signal voltage is lowest, and vice versa). I think that can be tweaked by adjusting the bias point - but you'd need someone other than me to help with that.
What was the rms signal voltage from the signal generator?
That said, the rounding is very smooth and looks pretty modest to my nooby eyes.
cheers, Derek