There is no such thing as weakened molecular covalent bonds. Unless, of course, you want to take a really contorted view of things. Heating the fuel (which is probably all they are really doing) will make the molecules expand. This does several things.
First, it reduces the density of the fuel. This will make the injector timings from the ECU be off by just a little bit so the injectors will put less fuel into the mix. The oxygen sensor feedback should let the ECU correct for this. But I doubt the old beast or a diesel they did their test on has this level of feedback.
Second, preheated fuel will take less additional energy to change state from a liquid to a gas. This means the fuel will cool the cylinder a little less and raise the operating temperature of the engine.
I can see how an oil snake preservationist could use the two facts to argue that molecular bonds have been weakened, but it's a real stretch. I can't say anything about what higher cylinder temps would mean to emissions, especially in a diesel, but I find their claim that you have to run the device for a few months to see the real benefits to especially unbelievable.
I just went back and looked at their fuel mileage data. The before data is all short mileage (94-122 miles). The after data is all long mileage (233-364 miles). If those two data sets represent identical use cases, I'll sell my half of the moon to the martians.
tl;dr; My considered opinion is that all their device does is preheat the fuel. Whether or not that actually makes any difference in fuel consumption and emissions is beyond my engine knowledge. The real question is do they know their data is crap?