I stuck with it longer than most people would with a bit of patience, and stopped assuming the obvious answer was right. The variable that mattered was the amount of information visible before matching — it turns out to matter far more than the interface. The gap was widest to a degree that mattered across the whole period I tracked.