In my view we need to consider two aspects. 1. Se it self is temperature dependent, and this dependence may be quite strong. 2. Thermal looses to surroundings. Both these aspects can be taken care -off by keeping temperature difference small ~ 5K.

Please check this link:

https://books.google.ca/books?id=VvCb_deT4kIC&pg=SA26-PA5&lpg=SA26-PA5&dq=Why+should+the+temperature+difference+be+minimized+during+seebeck+effect+measurement?&source=bl&ots=wCv_Ma865b&sig=fEcN8G0Rtp-4rau8rs3Oh7Go7Sg&hl=en&sa=X&ved=0ahUKEwjR9q-NorfKAhVEvIMKHX2MAJQQ6AEIRzAH#v=onepage&q=Why%20should%20the%20temperature%20difference%20be%20minimized%20during%20seebeck%20effect%20measurement%3F&f=false

Thermoelectrics Handbook: Macro to Nano, section 26.2. 

You will find your answer.

In my view we need to consider two aspects. 1. Se it self is temperature dependent, and this dependence may be quite strong. 2. Thermal looses to surroundings. Both these aspects can be taken care -off by keeping temperature difference small ~ 5K.

Be careful of convection losses