1. In merton's model, the interest rate just follows a brownian motion with drift. So it is normally distributed. We know that normal distribution can take the value from negative to positive infinity.
2. The mean and variance of CIR model is just the standard way to solve this kind of SDE. You first applying Ito lemma to the 'discounted process'. Say CIR model is like this: dr=a(b-r)dt+c sqrt(r)dz. calculate d(exp(at)r)first. This will kill the r in the drift term. So you integrate the equation and take expectation and z term will disappear, so that you get the mean. For variance, it is basically the same thing. What I recommend is you calculate E(r^2), and use E(r^2)-E(r)^2 to get the variance.
If you have any further questions, please let me know.
best