My (admittedly pretty hazy at this point) memory of heat engines is that their theoretical peak efficiency depends on the thermal delta they manage to achieve. Exactly the same resource that thermoelectric materials scavenge (albeit at miserable efficiency) into electricity.
Anybody who actually has some grasp of the matter want to chime in on where and why you would use thermoelectrics (and how efficient they would have to be) rather than simple insulation or one of the various waste-heat-recovery systems that transfer some amount of the heat remaing in outgoing exhaust gases into incoming working fluids?
Is the thermoelectric advantage purely that, assuming material reliability is OK, they are a 100% solid state, trivial to scale from 'handle with tweezers and magnification' to 'pretty large', and their output is easy to transfer and useful for all kinds of things after just a little DC-DC cleanup, or are there actually situations where they might be absolutely more efficient than insulation and heat recovery, rather than just easier to tack in almost anywhere in a design that you have a few extra cubic centimeters and expect a temperature difference?