In 184 adults with mild insomnia, 50 mg at bedtime raised self-reported sleep efficiency over two weeks. The same nights' recordings found no difference on any sleep measure.
The trial was powered on the diary outcome, so a flat recording is a difference not detected rather than a difference shown to be absent. Mild insomnia in adults averaging 44, not the older people the warnings are about. The drug arm ran 14 nights, so nothing here reaches beyond a fortnight.