When calculating delta-hedged option returns, the rebalancing frequency of the position in the underlying security dramatically affects the magnitude of the observed returns in the presence of microstructure noise. We show in a simulation environment that the mean bias of a delta-hedged return compared to an error-free benchmark increases with rising hedge frequency as a result of that noise. Standard asset pricing tests falsely identify significant option-return premiums for the illiquidity of the underlying asset. Because bias adjustments taken from the existing literature do not adequately address this issue, we suggest a new simple adjustment method using lagged hedge ratios to rebalance the position in the underlying.