This is an excellent point that made me think. Perhaps you could examine through how the data is collected. You could hypothesise that the mean age of death is

This is an excellent point that made me think. Perhaps you could examine through how the data is collected. You could hypothesise that the mean age of death is affected by non-testosterone increased aging such as murder by a jealous lover. These would pull down that average, where as the mode might show the more the 'typical' lifespan. However you might also hypthesise that high levels of testosterone might cause you to engage in behaviors that would make make other people want to kill you! Also including other risky behaviors such as driving fast this could also be a direct effect of testosterone. A very interesting thought!