Why aren't the strong LLNs and CLT contradicting each other?

The problem is that you left out the scaling. The average does converge to a constant: what converges to a normal distribution is $$\left(\text{average} - M\right)\sqrt{n}$$