Having calculated the power of my gas stove, I boiled a litre of water another three times. The time taken got longer and longer. This must be to do with lower pressure reducing the gas flow rate. I assume that a part empty cannister has lower internal pressure than a full one and adiabatic cooling also reduces the pressure. Here's the data starting with my first go:
Second reading is 33% bigger than the first.
16% bigger than the previous.
