Adaptive Traffic Signal Control (ATSC) systems depend on sensor readings in order to dynamically respond to the real-time traffic stream and update signal timings. In practice, sensors do not represent the state of the intersection with complete fidelity, and may not gather information needed for optimal or even effective control. The capability of an intersection-level controller also depends on its effective detection range, i.e. the range over which the intersection receives information about the state of the traffic network. Longer effective detection range may improve the controller, assisting with tasks such as progression formation. In this paper, we begin by making an analysis of the partial-observability (PO) of a fully decentralized regional ATSC system and clarifying the PO caused by sensor limitations and short detection ranges. We then construct a decentralized controller eMARLIN+ and experimentally verify on both representative synthetic and real-world scenarios that our eMARLIN+ effectively compensates for the two sources of PO, promotes coordination, and outperforms strong baselines in minimizing intersection delay.