Given the increasing popularity of 500 mL and 1 L units of bottled water, and the ubiquity of the 2 L HDPE beverage bottle, I think it would have been safe to measure the potable water output in liters. About the only things still sold by whole gallons any more are milk, water, iced tea, and lemonade, in the 1 gal jugs, or water in the 5-gal water-cooler jugs.
I'm more confused by the efficiency numbers exceeding 100%, which seems wrong. A theoretical efficiency of 100% would find the solar irradiance energy of 1 m^2 and then find the volume of fresh water per hour that produces an equal amount of energy when its salinity is increased to the mean salinity of seawater (about 3.5%). So if you can get 5000 Wh/m^2/day from insolation, and the energy difference between salt and fresh water is 0.810 W * h/L, 100% efficiency would be a 1 m^2 area producing 6173 L/day of fresh water. You're just dividing the daily energy of sunlight on your panel in Watt-hours by that 0.810, to get L/day.
The units in the article are all wrong anyway. They say "a rate of 5.78 liters per square meter", but there is no time factor mentioned whatsoever. An MIT roof gets mean 4.59 kWh/m^2/day of solar energy, so 100% efficiency would be 4590 * 1000/810 = 5667 L/m^2/day (1 m^3 = 1000 L). If the number given was per day, that's 0.1% efficient. If it's per hour, that's 2.4% efficient.