Spontaneous brain activity involves temporally fluctuating dynamic patterns across anatomically connected regions, yet conventional resting-state functional connectivity (FC) analyses based on blood oxygen level-dependent (BOLD) signals typically neglect underlying anatomical structure, when investigating temporal dynamics. To address this limitation, we propose a novel structural connectivity constrained BOLD (SCC-BOLD) framework that integrates anatomical pathways into functional time series estimation. Specifically, we use a graph-constrained regression model to estimate SCC-BOLD signals that explicitly reflect structural connectivity (SC). We then compute the SCC-BOLD-based time lag (STL) matrix and compare it with the intrinsic neural timescale (INT) to evaluate its biological plausibility. Using data from 982 participants in the Human Connectome Project database, we observe that the SCC-BOLD and original BOLD signals show similar but distinct patterns ( \(r = 0.625 \pm 0.108\) ). The mean inter-regional time delay across the brain, based on the SCC-BOLD, is 2.25s, significantly shorter than the 38.38s observed with the BOLD. Moreover, the correlation between STL and INT ( \(r = 0.429\) ) outperforms that of the BOLD-based TL ( \(r = 0.333\) ), indicating the improved biological plausibility of the proposed framework. Our results suggest that explicitly incorporating structural constraints into resting-state functional magnetic resonance imaging provides a principled approach for understanding the temporal architecture of brain function.