I'm not sure how many people know how an ELFI tester works and how many don't etc, but I expect its probably relevant here.:
Basically it creates a fault to earth through a resister (23A is a common figure), it measures the supply voltage with the fault there and without it there and sees how much the voltage sinks when it is there, ohms low can then be applied to find supply impedance.
It doesn't work perfectly:
1) As has been said, if your earth is missing, dangerous voltages will appear on metalwork, but as the tester is doing it for controlled durations, etc its likely to be less dangerous than if a real fault to earth occured
2) also the way of measureing it isn't perfect, other loads on the DNOs LV system stop you getting a perfect result (if you are feeling technical then:
http://www.theiet.org/Forums/forum/...=5&threadid=2600&highlight_key=y&keyword1=30% may be worth a read), the instrument accuracy is not that that amazing either
3) The 23A test will trip RCDs (and some B6 breakers) the solution to this is a much smaller current test in most machines on 'non trip' setting(although some do 'jam' the RCD with DC currents), the smaller the test the less accurate the result will be, in the case of RCDs the saving grace is Zs arn't quite as critical when an RCD is fitted, better to use a 23A test when no RCD though, also if you have a dodgy earth held on my not much, 23A might blow it clear and show up the problem where a smaller current will not, the opposite is true, it might fix a dodgy connection, if there is high resistance connection due to oxidation, a 23A test might clear it up, while a lesser current would not
4) The accuracy also drops off as the Zs decreases, basically the smaller voltage change the unit measures, the lower the accuracy (
http://www.theiet.org/forums/forum/...eadid=12820&highlight_key=y&keyword1=accuracy )