What is the test for "likely" in assault and abuse cases?

The People argue that the jury verdicts confirm that the jury did not interpret "likely" as meaning "possible." As they would have it, "The jury convicted appellant of abuse of an elder likely to produce great bodily injury but acquitted him of assault by means of force likely to produce great bodily injury. In the absence of record evidence to indicate otherwise, it must be presumed that the jury applied the same definition of 'likely' to each count. Had the jury believed that 'likely' meant 'possible,' it would have convicted appellant of assault by means of force likely to produce great bodily injury because it is possible to injure a person seriously by knocking him or her down on a concrete street. [] The jury's verdicts may be reconciled in such a way to shed light on its understanding of 'likely.' It plainly believed that assaulting a person with a single punch was not likely to produce great bodily injury because such injury was improbable. However, abuse of an 88-year-old elder by punching him in the face was likely to produce great bodily injury because his age and physical condition made such injury probable." While error may be found harmless based on other findings made by the jury (e.g., People v. Covarrubias (2016) 1 Cal.5th 838, 898-899), that is not the case here.