Using the AD8313, datasheet temperature coefficient is 'typical' 0.022dB/C.

In my circuit (measured with an opamp that follows and subtracts the offset and x6) I get an temperature coefficient of about 0,01dB/C (roughtly half the datasheet value) at 30-500MHz

I've tested a number of things and I'm quite sure it's not the conditioning circuit.

What is the expected variation of the AD8313 temp.co? Could this be effected on a lot/lot basis?

The temperature coefficient will vary over a statistical population, input power and as you said, over frequency. Please see Figures 6, 7, 8 and 9 in the datasheet.

I don't know what the exact expected range of variation of the temperature coefficient is for a population of devices, but I will say that the spread of devices, at 85C or -40C could be 1 or 1.5 dB. Plots 10 through 15 in the AD8318 datasheet show this behavior.