The Topaz II reactor used a thermionic conversion system which relies upon the thermal emission of electrons from a metal--very similarly to thermionic emissions of electrons in vacuum tubes. This poor efficiency is a reflection of thermionic emission--which is arguably better than thermoelectric generation (which is typically only about 0.5% efficient!)--however, it is an order of magnitude lower in efficiency than a good Brayton Cycle or Stirling Cycle heat engine. However, rotating and/or reciprocating machinery is more expensive, heavier and probably does not last as long as a thermionic generator (thinking in terms of what a Soviet Engineer may think...) Besides, if you're powering a military satellite, more power is not necessary to run cammeras, or ELINT gear. Higher power may help a radarsat--but going from a couple of hundred watts RF to a couple of thousand does not entail as much 'gain' as many people realize. The inclusion of a low noise pre-amplifier in the receiver can produce the same gain levels as increasing transmitted power (which is generally much more expensive.)

So the TOPAZ II was 'just' good enough to do the job.

For an ion powered spacecraft--much more power is needed--so I would expect that Brayton Cycle turbines would be the way to go.