It depends on the scope ... It will definitely clip the signal.
I had the same problem when doing the measurement. I used a real divide by two attenuator during the test.
The attenuator is 50 ohms both ways. Meaning it is 50 ohms towards the pulse generator and also matches the 50 ohms of the scope. (Plus we 'calibrated' the attenuator away.)
I guess I'll go with an attenuator to be safe.
The WaveJet that I have states that 5Vrms is a maximum for the 50ohm input but it goes up to 2V/div on 50ohms so presumably can show 16Vp-p so should be able to show a 10V pulse without clipping (I'm assuming that you can apply an offset of around -5V.)
SMA attenuators appear to be much higher frequency than BNC ones (18GHz vs 1 or 2GHz) for the same sort of price but I guess it makes little difference as the SMA to BNC adapter is probably only good to a GHz and my scope is 350MHz anyway.
(Hameg do a reasonably priced set of 4 attenuators and a 50ohm pass through rated at 1W and 1GHz with BNC connectors designed for their scopes.)