The size part may not be that bad. Chances are that much of the current version is radiation shielding. A larger cell would not need that much more shielding.
Some 8.8 % conversion efficiency sounds realistic. One may not reach that with silicon cells, but diamond may be a little better here, due to the high bandgap.
The loading curve would likely be very close to a PV cell: a constant current shunted with some diodes in series. With diamonds large bandgap they may get away with just 2 or 3 diodes in series.
A problem with the 1 W cell would be the heat : it would still produce some 10 W of heat all the time.
Even though a low energy beta emitter, there are limitations on handling radioactive materials. Even the 100 µW version may get problems with the legal limits and would likely need special care, tracking.
For the lifetime there is the decay of the radioactive source, but also the radiation damage to the conversion cells. With the low power density this may not be that bad.
There were some pacemakers with a nuclear battery (AFAIK still the thermolelectric way with PU238) made in the USSR.
When comparing to other batteries the nuclear battery would provide power all the time, a normal battery one uses it's capacity when actual current flows