Thanks for the thoughts - I'm comfortable that my reference temperature is as good as I can get.
Although I'm using a Lascar professional temperature logger inside the VP2 ISS, I actually have two other reference sources - a mercury thermometer (±0.5C) in a meteo cage positioned underneath the ISS and a calibrated Fluke platinum thermocouple (±0.2C), both of which I've been checking the Lascar against.
However I've really been using the Lascar (±0.25C with 0.5C internal resolution) as a proxy, as it's the only device which I can actually fit next to the ISS sensor,
Also, main comparisons have been made when temperatures have been reasonably stable for at least an hour, so there shouldn't be any response time issues.
I'm definitely puzzled.