Statistics for Semiparametric recurrent event vs time-to-first-event analyses in randomized trials: Estimands and model misspecification