My experience over a number of vehicles has been that the highway figure is pure fantasy and that the city figure is a reasonable representation for what I can expect in daily driving. This continues to hold true for both my '07 Vibe (05spd) and '09 Montana (extended).
For strictly city stop-n-go driving, I figure the sky's the limit as I have been stuck in traffic that resulted in a 35 km trip taking 3-1/2 hours. There is no such thing as fuel economy in those circumstances, unless you're driving a hybrid. But it's still better than being jammed into a TTC bus.
The highway driving figure is not fantasy, I get it in summer on the highway driving at 100 Km/h, avoiding bursts of acceleration. It helps that I have a Scangauge that is a very good indicator of instantaneous fuel consumption that can be adjusted depending on the results at the pump, so I know how to maximize fuel economy.
People just think it's fantasy because they do not drive in a conservation-minded fashion. In stop-and-go traffic, you will never see it, and when they finally get on the highway, they drive like bats out of hell at 120-130 Km/h and then wonder why they get bad fuel economy. Hint: driving at 120 Km/h will burn around 20% more gas than at 100 Km/h*, your engine or transmission doesn't matter much in this, it's air resistance and simple physics... unless you simply drive behind another vehicle and use their slipstream, which need not be dangerous, just match speed and stay 2 seconds behind them.
The way I see it, the Canadian fuel economy for the highway is representative at what your car gets at 100 Km/h constant speed on a flat highway with no wind. The city rating is closer to a 50% suburban, 50% highway commute, so most people should expect that number in their commute. If you're in a very urban setting, then all bets are off, I suggest looking at US city rating for that.
*See
http://www.mpgforspeed.com/ for info.
Oh, and I support the comment about the very poor methodology of the 50-litres test. There was no rigorous accounting for any variation in tank size between the different vehicles, driving a car until you think you've burnt 5 litres of its 55-litres tank is horribly unprecise, and I noted that cars that had LESS than 50 litres in their tank size had been presumed to have 50 nonetheless, like the G5 or Cobalt, who have 49 litres but who had their fuel consumption calculated on the assumption of a 50-litre tank.