The main issue is that they're really rather inefficient. The efficiency of the thermocouple at converting thermal->electrical is only about 5-10%, and combined with the cost of shielding, expense of the radioisotope to begin with, and security/safety considerations, they're really only suitable for niche aerospace/defense applications.

Edit: I was curious if anyone had considered a stirling or other heat-engine driven by decay heat, and found https://en.wikipedia.org/wiki/Stirling_Radioisotope_Generato... which looks like it can hit 20+% efficiencies. The downside is that unlike thermo-electric/Seebeck effect converters, they have moving parts that could be a threat to reliability, which is the major issue when you're a planet away from the nearest repair tech.

[1] https://www.orau.org/ptp/collection/Miscellaneous/pacemaker....