As a real-world illustration of this point, one of the Kickstarter projects I backed recently is for a lantern powered by a candle. It uses a burning candle as a source of heat, uses the heat to power a thermoelectric generator, and uses the electricity to power LEDs.
Huh? Why do that, instead of just using the light from the candle? Especially given that thermoelectric generators are only about 5% efficient, and a lot of the heat is going to escape before even getting to the generator, and LEDs are less than 50% efficient themselves. Well, it turns out that a burning candle flame puts out about 80 Watts of heat, and even with all those inefficiencies, you can get about 130 lumens from the LED. That's about 10 times as much light the candle flame itself puts out.
That tenfold increase, at very low power levels, is worth the complexity and capital cost of the unit, assuming (which is the risk I'm taking as a Kickstarter backer) that the unit works reliably as designed. It will be a useful part of my power outage (medium-term off-grid living) kit. Candles are way more expensive per kilowatt-hour than grid power (about $2 vs. about $.10), but not nearly as expensive as single-use batteries, which can easily cost $100 or more per kilowatt hour.
Would I buy a device requiring a reaction cell, lots of nickel tubing, some potassium carbonate, and some heating elements, that would produce a mere 90 steady usable Watts from water? As it happens I'm pricing out battery and solar panel options to achieve just about that level of steady power availability (2.1 kilowatt-hours per day) under bad-case conditions (e.g. overcast winter days), because that's a threshold level for keeping a small refrigerator going, along with a few hours a day of Internet access and all the USB device charging I would need. The cost for that capability ends up well over $5,000; more if you want the system to last under steady use (requiring a lot of additional battery capacity to avoid deep-cycling the batteries). And the system will still require a lot of care and maintenance to keep it ready for use, and even more when in actual use.
What was in that fully operational mid-90's Mills design, that would have cost more than half that (allowing for a retail markup) to mass-produce? Some nickel tubing? No.
For comparison, consider what a gas-burning home hot water heater requires. A glass-lined vessel that holds tens of gallons under pressure, a high-temperature combustion chamber, electronic systems for temperature control and safe condition monitoring, a thermocouple to power the electronics, pressure safety valves, a magnesium sacrificial anode, copper pressure fittings, and more. Yet they don't retail for anything close to $5,000.