IR works reasonably well if you know the emissivity of your target very well, or you use extremely expensive dual-band IR - even then, its nothing like as good as a good T/C. It doesn't work well enough on liquid metals to satisfy my customers, who refuse IR as primary measurement for temperatures in my instruments - they insist on type C TC and they are horrible damned things - a sniff of oxygen and you need a new sensor. If your precision is <=0.2 C, I can't really see an IR working at all well.

Steve