MicroRNA (miRNA) has been mostly associated with decrease in target protein expression levels. Recently, `unexpected' observations of increase in target protein expression attributed to microRNA regulation have been reported. We formulate a comprehensive model for regulation by miRNA that includes both reversible mRNA-miRNA binding and selective return of RNA. We use this mathematical model incorporating multiple individual steps in the regulation process to study the simultaneous effects of these steps on the target protein level. We show that four dimensionless numbers obtained from 12 rate constants are sufficient to define the relative change in steady state target protein levels. We quantify the range of these numbers for which such pleiotropic increase in protein levels is possible, and interpret the experimental findings in the framework of our model such that the results are no longer unexpected. Finally, we show through stochastic simulation that the nature of the target protein distribution remains unchanged and the relative steady state noise levels are also completely defined by the values of these dimensionless numbers, irrespective of the individual reaction rate constants.