It looks like the temperature for the resistor in 'port' remains fixed at 290K. That will give you incorrect input noise power and consequently incorrect noise factor and NF if you happen to measure NF at temperatures other than 290K. I have confirmed that by connecting a port (Rs) to a noiseless shunt resistor (Rs) and VCVS with a gain of +1 and an output resistance of noisy Rs. For this circuit, theoritically, the F is exactly 5 at any temperature. But noise analysis in spectre shows F as increasing with temperature and exactly 5 only at 290K. Does anyone know how to set the port so its temperature varies with the circuit temperature?

yes the IEEE standard NF is that the source is at 290K. (that only because then kT=4e-21 which is easy for pre-calculator days of 1957...)

if you are using port in spectre (analogLib/port) you can define in noise temperature in the form.