Interval-censored time-to-event and competing risk with death: is the illness-death model more accurate than the Cox model?

If follow-up intervals are wide and the exposure has an impact on death, then the illness-death model for interval-censored data should be preferred to the standard Cox regression analysis.