You know how people say you should trust hobby grade water test kits, well the same can be said for cheap power meters, for certain types of load they read hopelessly wrong. Probably fine for electrical heater or normal tungsten filament light bulb or a kettle but may be hopelessly wrong for anything with a large motor in (try vacuum cleaner) or anything will electronics in, especially cheaper items.
Its all to do with power factor, your power meter does not take into account power factor.
Power factor - Wikipedia, the free encyclopedia
I have a more expensive power meter (£20, Maplin ?) and it also reads power factor, enabling me to correct the power reading. Maplin N47KB reads real power and apparent power.
I have a set of PC speakers, which reads 20Watts when in standby (this is apparent power), however the power factor is a very poor 0.2, which gives real power (ability to work, energy/second) of 20 x 0.2 = 4Watts, which matches the manufacturers quoted value in standby.
You also get some other weird readings like my 10year old PC measures 140W when "doing nothing" and stays at about 140W when maxed out at 100% CPU, despite obvious heat being generated. However when idle power factor is 0.35 (50Watts in idle) going to 0.95 at 100% CPU.
So the real power taken by your lights when low, is liable to be significantly less than the 20W your power meter is reading (I may be wrong of course).
Oh your domestic electricity meter charges you real power (the lower value), so your 20W value is again meaningless. Industry gets charged apparent power, so they go to significant lengths to ensure power factor is as near to 1.0 as possible.