Context. We analyzed spectropolarimetric data recorded with Hinode/SP in quiet-Sun regions located at the disk center. We found single-lobed Stokes V profiles showing highly blue- and red-shifted signals. Oftentimes both types of events appear to be related to each other. Aims: We aim to set constraints on the nature and physical causes of these highly Doppler-shifted signals, as well as to study their spatial distribution, spectropolarimetric properties, size, and rate of occurrence. Also, we plan to retrieve the variation of the physical parameters with optical depth through the photosphere. Methods: We have examined the spatial and polarimetric properties of these events using a variety of data from the Hinode spacecraft. We have also inferred the atmospheric stratification of the physical parameters by means of the inversion of the observed Stokes profiles employing the Stokes Inversion based on Response functions (SIR) code. Finally, we analyzed their evolution using a time series from the same instrument. Results: Blue-shifted events tend to appear over bright regions at the edge of granules, while red-shifted events are seen predominantly over dark regions on intergranular lanes. Large linear polarization signals can be seen in the region that connects them. The magnetic structure inferred from the time series revealed that the structure corresponds to a Ω-loop, with one footpoint always over the edge of a granule and the other inside an intergranular lane. The physical parameters obtained from the inversions of the observed Stokes profiles in both events show an increase with respect to the Harvard-Smithonian reference atmosphere in the temperature at log τ500 ∈ (-1, -3) and a strong magnetic field, B ≥ 1 kG, at the bottom of the atmosphere that quickly decreases upward until vanishing at log τ500 ≈ -2. In the blue-shifted events, the LOS velocities change from upflows at the bottom to downflows at the top of the atmosphere. Red-shifted events display the opposite velocity stratification. The change of sign in LOS velocity happens at the same optical depth in which the magnetic field becomes zero. Conclusions: The physical mechanism that best explains the inferred magnetic field configuration and flow motions is a siphon flow along an arched magnetic flux tube. Further investigation is required, however, as the expected features of a siphon flow cannot be unequivocally identified.