Briffa et al. (1998), https://doi.org/10.1038/35596 published a seminal paper on the reduced sensitivity of annual tree growth to temperature across Northern Hemisphere treeline stands. By averaging tree-ring chronologies to sub-continental means, they found decade-long trends in maximum latewood density (MXD) progressively diverging from observed warming temperatures since the 1960s. This divergence challenges the reliability of the proxy, but the lack of an intercontinental network of up-to-date MXD chronologies extending into the 21st century hindered large-scale evaluations of the phenomenon, leaving it unresolved. Here, we introduce nine new MXD chronologies along the North American Rocky Mountains between 38° and 69°N and analyze their trends after applying novel approaches to preserve low-frequency variability. When following the original Briffa et al. (1998), https://doi.org/10.1038/35596 methodology, the divergence between increasing temperatures and MXD chronologies reaches offset values greater than 1°C by 2020 CE. However, divergence markedly decreases and even disappears entirely when MXD chronologies are (a) based on high-replication data sets including differently old trees, (b) detrended using signal-free age-dependent splines instead of Hugershoff curves, and (c) calibrated against optimum season instead of April–September temperatures. MXD chronologies north of 60°N exhibit stable relationships with regional summer temperatures on interannual to multi-decadal timescales, but at the southern sites, a lack of high-frequency proxy-target coherency is evident starting in the second half of the 20th century. This study emphasizes the importance of careful site and target selection, sampling design, and chronology development for overcoming the divergence problem and reconstructing summer temperatures from MXD data in North America.