These associations showed a general pattern of higher hazard ratios with higher noise exposure, but with a levelling off or even small declines in risk at higher noise levels

I wouldn't be surprised further study will show the last bit is due to earplugs at higher noise levels, resulting in a relative reduction in sleep disruption. Unless they also took earplug use into account already, then I'm genuinely curious what could be the reason