Aillio needs to answer here but in the interim here’s my take on your 3 questions-
1- If IBTS < Bean Temp, odds are very good the IR sensor of the IBTS is dirty. If you look at the sensor with a fiber-optics probe you (probably) won’t see anything on the IR sensor lens, but cleaning the lens with a cotton swab moistened with alcohol (several times!) will improve or eliminate the error. If you roast dark, IBTS will probably collapse down toward Bean Temp after 1C but shouldn’t cross below.
2- Currently RoR is based on Bean Temp only, though I understand the IBTS-based version will be coming at some point. The issue is artifacts in IBTS data (from IR sensor sample rate vs. bean and drum movement) that cause the data to move around a lot. Since RoR represents the change in temp data vs. time, Bean Temp-based RoR will be more readable than RoR based upon IBTS… at least at this point.
3- Bean Temp is thermocouple-based. Thermocouples have been around a very long time and generally are durable and repeatable within their usable range. But their response is slow because of thermal mass of the thermocouple, the mass of the probe case and the surrounding metal that has to be heated. On the other hand, the IR data from IBTS contains artifacts that make reading it a challenge. And then there’s the issue of IR calibration errors from surface contamination on the lens interfering with IR emissions (1- above).