Past forest mortality after disturbances creates one source of uncertainty that needs to be taken into account to reconstruct biomass dynamics. Additionally, using unitless normalized data from tree-ring records to calibrate stand productivity in vegetation models constitutes another source of uncertainty for model calibration. In this presentation I would assess these two sources of uncertainty and discuss the influence of calibration technique in model-data fusion. I combined data from permanent resampling plots and dendroecological plots to estimate annual woody biomass growth (ABI) in several forests. ABI were used to benchmark a process-based vegetation model to analyze time variability in forest productivity and carbon allocation. The model used implements source and sink limitations explicitly. Disturbances, species life-history strategy and climatic variability modified the carbon allocation pattern. Bias in tree-ring reconstructed ABI increased back in time from data collection and with increasing disturbance intensity. ABI bias raised to over 100% in stands with major disturbances. Environmental variability and leaf area explained much variability in woody biomass allocation. Divergence between tree-ring estimated and simulated ABI were caused by unaccounted changes in allocation or misrepresentation of some functional process independently of the model calibration approach. Using directly unbiased estimates of biomass growth improved model performance. Higher disturbance intensity produced greater modifications of the C-allocation pattern. Legacy effects from disturbances increased error in reconstructed biomass dynamics, reducing the potential use of ABI as a proxy to net primary productivity.