That is because for categorical dependent variables, the sample statistics for model estimation are correlations not variances/covariances. If you treat the variable as continuous, you will obtain a variance like you get in Stata.
Yes!, now I have the variances and they are pretty the same that I get in Stata.
Just to be reassured: I must take them away from the CATEGORICAL list only for the Basic analysis, right? I understood that when I run the model, the categorical dependent variables must be in the list.