Due to progress in high-temperature superconductor (HTS) technology, research in the application of superconducting quantum interference devices (SQUIDs) in time-domain transient electromagnetics (TEM) has intensified. Several TEM systems using HTS SQUIDs have been developed and tested in numerous field trials. In this paper, the reliability of SQUID TEM data is investigated by comparison with commonly used induction-coil data. Generally, a good agreement between SQUID and coil TEM data was found. However, two effects were observed with the SQUID TEM system which were not visible in coil measurements. The first effect is a "frequency dependence": the response signal measured by the SQUID depends on the base frequency of the transmitter signal, decaying faster for higher frequencies. The second effect is a "zero-crossover": the SQUID response data becomes negative for late times. We propose a model for magnetometer TEM measurements that provides a possible explanation for these two effects. This model takes into account remnant responses induced in the ground by repetitive transmitter signals. Our simulations of coil and magnetometer responses show that, at moderately conducting ground, the influence of remnant responses is negligible in coil recordings, whereas it leads to a typical frequency dependence of magnetometer recordings. We show also that the influence of remnant responses may lead to the sign reversal of magnetometer data over highly conducting ground structures. To optimize magnetometer TEM measurements, it is advantageous to use lower transmitter frequencies than typical frequencies used for induction coil measurements. A recalculation of the original ground response on the basis of the proposed model may also be of advantage.