I'm thinking the roof top was the best bet. It's still the wrong helio for training unless you already have a cert and are now training to fly a R22 and have been thru SFAR 73. Looking at the ground for an auto has some advantages like more time to plan out the thing, maybe more room for a better flare, more room for ground run out. Its also true there would be lower density alititude at GL than 700 or so AGL but in city like LA I would'nt count on having to many places on the ground that are perfect for an auto let alone be lucky enough to be near one when it happens. Main and/or tail rotors stricking pedestrians or vehicles or even something else first like a sign post then fragmenting is'nt good on a crouded city street. The majority of the rotor fragments (if not all) probably stayed on the roof. The saftey nets and shelf on a FAA approved roof top pad exist for a reason. Another thing as that going down around all the buildings would make for some unpredictable winds. I have become good at predicting/avoiding structure generated shears by observing the wind sock and looking at the surrounding structures, terrain and trees and extrapolating their effects. I have watched a EMS helio land at a local hospital which had 3 wind socks all pointing in different directions. One was on a pole at GL and two on two different levels and points on buildings. Think high, middle, ground level. He had to decend thru two shifts in wind direction and velocity. Scrubing altitude off at the rate of an auto thru unpredicable winds while traveling at the recomended auto rotation airspeed per the POH is'nt a good idea in downtown LA.