Brillouin optical time domain reflectometry sensor utilizes a peak frequency in a Brillouin gain spectrum to measure distributed strain and/or temperature. When the strain and/or temperature changes significantly within a spatial resolution at the measuring position, the measured Brillouin gain spectrum has multi‐peaks with competing intensities and the noise overlapping may cause a big jump to a falsely determined frequency. Such a rare spot has been overlooked but are found by recent sensors with a higher sampling resolution. The signal‐to‐noise ratio improvement or multi‐peak fitting methods is effective, but it cannot eliminate this problem in essence. In this study, an analytical method is developed to correct the erroneous peak determinations based on the nature of the spectral‐intensity trends along with the distance axis. When a peak frequency transits from one peak to the other, the intensity of the former peak decreases and that of the later increases, and those levels intersect. Simulations are conducted with a spatial resolution of 1 m for frequency transitions with a frequency difference of 3–300 MHz. For differences larger than an intrinsic Brillouin bandwidth, this method is effective in avoiding a large jump to a false peak. © 2022 Institute of Electrical Engineers of Japan. Published by Wiley Periodicals LLC.