The MARVELS (Multi-object APO Radial Velocity Large-area Survey) project will monitor the radial velocities of 3300 stars. As a result, the project is likely to find rare objects, such as brown dwarf (BD) companions in the so-called brown dwarf desert. Two BD candidates in the desert have already been uncovered by the survey. Until recently, MARVELS Candidate 10 (MC10) appeared to be yet another candidate for the desert. The RV data seemed to neatly fit an eccentric orbit with a 237-day period and a minimum mass of 50 Jupiter masses. However, there were two anomalous points in the RV follow-up, which were initially rejected because they were thought to be a result of changes with the instrument. Then it was discovered that the original MARVELS data possessed similar anomalous points which had also been excluded as outliers. A thorough re-examination of the cross-correlation functions (CCFs) in both data sets revealed that the two anomalous points in the follow-up data had double-peaked CCFs, and the MARVELS data showed CCFs that seemed too broad compared to the other MARVELS candidates. This led to the conclusion that MC10 is most likely an eccentric spectroscopic binary with the long axis aligned nearly perpendicular to the line of sight. Therefore, it is only for a brief moment near periastron that the two radial velocities are large and disparate enough to be resolved as a double-peaked CCF. This cautionary tale implies that RV searches for substellar companions must ensure full phase coverage, and must carefully examine how the CCF changes with time, in order to avoid this kind of astrophysical false positive.