Adaptive traffic signal control can reduce congestion, but network-wide averages may conceal access-level penalties and spillback. This study evaluates an interpretable Coordinated Adaptive Control Model (MCAC) for two closely spaced signalized junctions in Loja, Ecuador. MCAC uses eight lane-area detectors to trigger temporary coordinated decongestion phases, inserting additional signal-control time before resuming an observed 137 s fixed-time base plan. A matched stochastic SUMO 1.27.0 experiment compared fixed-time, MCAC, and SUMO native actuated control at 80%, 100%, 120%, and 140% of nominal demand, with 10 common-random-number replications per condition (120 runs). The study is a controlled comparative simulation rather than a field-calibrated prediction of absolute corridor performance. Relative to fixed-time control, MCAC reduced mean waiting time by 51.1%, 38.7%, 36.2%, and 37.0%, respectively, with Holm-adjusted Wilcoxon $p=0.0234$ at each demand level. MCAC also reduced extreme waiting, departure delay, and a supplementary combined measure of pre-entry and in-network stopped waiting. Benefits were unevenly distributed: Orillas del Zamora improved strongly, both Guayaquil directions incurred systematic waiting penalties, and Nueva Loja shifted from a low-demand penalty to a high-demand benefit. The generic independently actuated benchmark exhibited prolonged blocking and reached 164 teleports at 140% demand; the data do not isolate the specific downstream-blocking mechanism or establish a consistent teleport advantage over fixed-time control. These comparative simulation findings support joint reporting of aggregate efficiency, access-level redistribution, and operational robustness, without implying field-validated effectiveness.