Doesn't the uncertainty principle just set a lower limit for how well a perfectly skilled pencil balancer would be able to balance it though? As in, if the experiment is repeated enough times, by pure chance \theta_0 \dot\theta_0 could at times be smaller than \frac{3 \hbar }{2 ml^2} ?