In realistic graphical user interfaces, clicking outside a target would require recovery time from an error, e.g., selecting an unintended hyperlink requires reloading the previous webpage. Several studies on target-pointing tasks have examined the effects of a “penalty time” for mis-clicks on movement time and error rate, but the effects on throughput (i.e., a unified metric on pointing performance) have not been thoroughly investigated. We conducted a crowdsourcing study with 127 workers and a lab-based controlled study with 30 university students. The penalty times varied from 0 to 10 s, and the results consistently indicated that the throughput differences were less than 5%, although the error rates were remarkably different when the penalty time was 0 s. This demonstrated the potential of normalization capability of the effective width method of Fitts’ law, and the throughput is considered a valid metric when researchers would like to compare several task conditions in realistic user interfaces in which error operations induce different recovery times. However, because the model fitness using the effective width method was comparatively low when the penalty time was 0 s, comparing throughputs for different conditions is not recommended. We also discussed potential issues related to the effective width method of Fitts’ law, such as endpoints not following a linear relationship to the given target width only under the zero-penalty condition.