You get some zero estimates because there aren't enough degrees of freedom; in other words, if you have a class variable with 4 levels, you can only estimate the coefficient for three of them, and the fourth is set to zero and it's variability cannot be estimated so the variability is set to missing.
You should be looking at the least squares means to help you understand the model, instead of the model coefficients. The least squares means do not get set to zero, and so are much more interpretable.
... View more