The nozzle efficiency and performance of a magnetic nozzle operating at low power (<200 W) are experimentally and analytically characterized. A suite of diagnostics including Langmuir probes, emissive probes, Faraday probes, and laser induced fluorescence are employed to map the spatial distribution of the plasma properties in the near-field of a nozzle operated with xenon and peak magnetic field strengths ranging from 100 to 600 G. The nozzle efficiency is found to be <10% with plasma thrust contributions <120 μN and specific impulse <35 s. These performance measurements are compared with predictions from quasi-1D idealized nozzle theory and found to be 50%-70% of the model predictions. It is shown that the reason for this discrepancy stems from the fact that the underlying assumption of the idealized model-that ions are sonic at the throat of the nozzle-is violated at low power operation. By correcting for the shifting location of the sonic transition point, the model and experiment are made to agree. The physical mechanism by which the sonic line moves with respect to the nozzle geometry is attributed to non-ideal behavior at low power. In particular, it is posited that the low ionization fraction at these low operational powers gives rise to neutral-collisional effects in the near-field of the device that can impede ion acceleration. The roles of ionization, enhanced electron resistivity, and charge exchange collisions are all examined. It is found that the ion sonic transition location is most correlated with the ratio of the charge exchange mean free path to the characteristic electrostatic acceleration length giving rise to an effective drag term on the ions.