LNA noise figure measurement - Y-factor vs cold source

Getting 0.3dB discrepancy between Y-factor and cold-source NF measurements on a 2GHz LNA. Which one do you trust? My noise source ENR cal might be the culprit.