For tracked observations, position switched observations results in a noise rms larger than frequency switched observations for the same elapsed telescope time. In other words, frequency switched observations are twice as efficient in time to reach the same rms noise than position switched observations.
However, time efficiency is not the only criteria of choice. Indeed, with the current generation of receivers (before march 2009), the IF bandpass is much cleaner in position switched than in frequency switched observations. Frequency switched is thus really useful only when the lines are narrow so that the IF bandpass can be easily cleaned out through baselining with low order polynomials.