I’ve seen the same thing. My current favorite suspicion is that we may be seeing some aging in the IR sensor as well as some accumulation of roasting ‘stuff’. ‘Stuff’ means I have never seen anything on the cotton swab after cleaning, yet cleaning makes a difference in IR levels during the roast. ‘Aging’ means that no amount of cleaning has restored the original values I saw before I realized there was a need for cleaning.
Recently I noticed that batch size changes the difference between BT & IBTS levels: a larger batch brings IBTS closer to BT values, especially at the end of the roast. Unfortunately I can’t tell if IBTS is lower or if BT is higher. This isn’t a change in IBTS behavior- it’s just that I always roast 550 gm batches till the day I tried one 800 gm batch and one 1 kg batch. The difference between IBTS vs BT was noticeably smaller. And again, there’s no clear way to tell if one or both of the sensors is different.