Right, so your citation is specifically density-impulse which is rather different to specific impulse, the conventional metric when discussing 'efficiency' of an engine, but of course you can define efficiency to be many things depend on the context, like thrust to weight ratio or whatever. But density impulse is quite a specific and generally not useful metric for rocket engines out side of e.g. a missile that has to fit its propulsion unit into very constrained packaging. For a conventional rocket, e.g. to launch a satellite, it is absolutely the case that adding mercuty would make the performance much worse. If you read the page before the one you cited, Clarke says as much:

I calculated the performance of Cavea A with various proportions of mercury —up to six times the mass of the primary propellant. (It was easy to fit mercury into the NQD calculation method.) As expected, the specific impulse dropped outrageously as mercury was added to the system, but the density impulse (specific impulse X propellant density) rose spectacularly, to peak at 50 percent above that of the neat mono-propellant with a mercury/propellant ratio of about 4.8.

Specific impulse is overwhelmingly the metric that matters for rocket performance in all but a small number of applications, so I feel justified in saying that your original claim that:

A traditional rocket becomes more efficient if you mix some mercury into the fuel. The loss of energy is made up for by the extra mass.

is simply not correct!