In the study of surface processes, it is generally assumed that erosion occurs equally throughout the soil profile so that chemical depletion of the topsoil can represent the intensity of chemical weathering and the duration of surface exposure to cosmogenic radiation can reflects the soil residence time, and then the rate of erosion can be calculated. In comparison with fresh bedrock, the depletion of soluble elements in soil mainly comes from fine-grained secondary clay components, while the depletion degree of detrital minerals is weak. The preferential erosion of fine-grained secondary clay will lead to the underestimation of weathering intensity, and the retention time of detrital mineral will be longer than the total retention time of soil, and thus the soil erosion rate will be underestimated. Based on the uranium isotope comminution ages of soil in the Lesotho Highlands, we found that erosion operates differentially between the detrital and authigenic components of the soil. Uranium isotope comminution ages show a soil residence time of (543±32) ka for the detrital particles. In contrast, soil residence time of the authigenic phases is constrained to be (22±11) ka according to the accumulation of recoiled 234U from the absorbed 238U to river water. The residence time of secondary clay matches with the regional erosion rate 24-33 t/(km2·a) calculated from weathering flux, indicating that secondary clay is the main component of soil erosion. The results indicate that the decoupled erosion of different components in soil may be common. This finding implies that the intensity of weathering based on bulk soil erosion and the rate of soil erosion determined by exposure dating of coarse soil grains may be invalidated due to the preferential erosion of authigenic particles. As a result, a lower estimate of weathering flux may be made, and therefore the role of chemical weathering in the global carbon cycle could be underestimated.