"Accidents in labs" can be:
-misinterpreted results
-some overlooked systematic error in the experiment
-rushed results
-deliberate hoaxes
-legitimate discoveries.
Whereas the first two are the most common, they are usually weeded out by peer review.
Rushed results are a special case of the two first, in which case the peer review system does not have the chance to verify the results and check and triple-check the experiment design for flaws.
Deliberate hoaxes are weeded out by peer review, but for a short time they may look like rushed results; the difference here is that experiment results have been tampered deliberately, where as in three other cases they are just flawed because of whatever reasons.
Despite all this, the fourth group
is very real as well as damned important for advances of science. For more examples than Galvani's frog legs twitching, I might mention antibiotics (failed bacteria culture), first observations of cosmic background radiation (initially suspected to be pigeon poo in the antennae), vulcanizing rubber, dynamite, nylon, X-rays et cetera.
Something being discovered by accident does not automatically make it inaccurate or nonexistant.
Although I'm not exactly holding my breath about this, it does feel more than little like Steorn (see the

-smiley in my first message to gauge my initial reaction), seeing how little they actually tell about energy input/output on the scope of the whole experiment - as was said, it doesn't help if the flames are hot if you waste energy in keeping them hot. Water does not release energy when it breaks into hydrogen and oxygen; it requires it.
And whatever the salt ions role in the equation is, I haven't got the faintest idea.
