On the determination of biases in satellite‐derived temperature profiles