Mars has seasons, just like Earth. While Mars weather is significantly less interesting than Earth's, due to the thinner atmosphere, the average temperature difference between summer and winter can be more than 50 C. The difference between daytime high and low can be more than 120 C.
If we consider the three forms of heat transfer, we can probably eliminate conduction as a significant factor, given that Curiosity sits high off the ground on six wheels. Although the wheels and suspension are all metal, along with the body, the small contact surface area and relative isolation of the RTG suggest to me that the thermal path to ground is pretty poor.
Furthermore, NASA itself says that the Mars atmosphere acts more like a thermal insulator than a conductor, because of the low density and thermal conductivity of CO2. Therefore, we can eliminate convection as a significant source of thermal variation.
Thus, we are left with the varying solar irradiance impinging on the RTG itself, as well as the portion of the rover body which can transmit a meaningful amount of heat to it as the only significant sources of thermal variation. Both the day/night and seasonal cycles will produce large swings in temperature, as cited above.
The "fin-root temperature" of the RTG is only 157 C. If we look at the average high temps over the year as measured by Curiosity itself, we see a variation of almost 30 C. At a daytime high of 4 C, we are looking at a maximum Carnot efficiency of about: 1 - (277/430) ~= 36% vs. -23 C, which gives: 1 - (250/430) ~= 42%. Obviously, the colder weather improves the theoretical efficiency by up to 17%. Thus, it is reasonable that the actual efficiency is affected by closer to 10%.
Thermodynamics (Added via EDIT)
Curiosity has two temperature sensors: one which measures air temp, and one which measures IR from the ground as a proxy for ground temp. As you can see from the Mars weather report, the ground temp is about 5 C higher, on average, than the air temp. Thus, the difference in air vs. ground does not seem to be a huge contributor (given that the surface rock of Mars serves as an enormous thermal mass, it makes sense that it retains heat better than the atmosphere, leading to the higher average temp).
Calculations have been conducted by Randall Osczevski to estimate the "Earth Equivalent Temperature" (EET) on Mars, which is basically a "wind chill factor" to compare what a human might perceive the Martian weather to be like. Due to the thin atmosphere, Mars would "feel" much warmer than the -60 C nominal temperature, which means that the atmosphere would carry away far less heat by convection. In fact, solar + ground radiation dominates the effective temperature. Transiently, the ground and air temps may differ by up to 20 C, just as air on earth can vary in temperature by a considerable amount just a few tens of meters above the surface.
The atmospheric pressure near the surface of the Earth is on the order of 1000 mbar, whereas, the comparable pressure on Mars is more like 6-8 mbar (more than 100x lighter). I presumed that the lower pressure would result in a lower thermal conductivity, but CO2 appears to be fairly insensitive to pressure after all. Rather, the chemical composition accounts for a 2x difference between Earth (18.5 mW/K.m) and Mars (9.6 mW/K.m).
It's hard to say what the exact contributions of the air temp, ground temp, and direct irradiance are to the fin temperatures of the RTG, but it seems reasonable to consider the ground as a kind of "solar concentrator" in that IR emitted from the ground due to solar heating will impinge upon the RTG, raising its temp somewhat. Any IR coming from the air is surely negligible compared to the ground, leaving the air contribution to be the convection value. Given the EET calculations mentioned above, it seems that even this effect is relatively small.