A rare dataset of in-situ 10Be from high-resolution depth profiles, soils, rock outcrops and stream sediments is combined with geochemical analysis and modeling of regolith evolution to understand the variability of denudation rates in a mountain watershed (Strengbach critical zone observatory). High-resolution depth profiles are key to detect the presence of mobile regolith and to highlight how it affects the critical zone evolution. The modeling of regolith evolution and 10Be concentrations along depth profiles allow to estimate both the cosmic ray exposure age (19 kyr) and the mean denudation rate (22 mm.kyr-1) of the regolith without any steady-state assumption on 10Be concentrations. Comparison with maximum denudation rates inferred from topsoil samples collected from the surface of the depth profiles and calculated using the temporal steady-state assumption of 10Be concentrations highlight an overestimation of denudation by a factor of two. Maximum spatially-averaged denudation rates determined from stream sediment samples also likely overestimate denudation rates by a factor of two. These biases are significant for investigating the geomorphological evolution and we propose a method to correct denudation rates using the inherited 10Be concentrations and the cosmic ray exposure age deduced from the high-resolution depth profiles. A key result is also that a steady-state of 10Be concentrations and a steady-state of regolith thickness are two different equilibrium states that do not necessarily coincide. The comparison between locally corrected and spatially-averaged denudation rates indicates that the watershed geomorphology is not in a topographic steady-state but is modulated by regressive fluvial erosion. Nonetheless, our study demonstrates that even in a watershed where the steady-state assumption of 10Be concentrations is not verified, the spatial variations of in-situ 10Be concentrations in sediments still carry qualitatively relevant information on the geomorphological evolution of landscapes.