In many applications in social and behavioral sciences, measurement scales of variables are arbitrary. Although it should not be viewed as a universal solution, some researchers resort to the standardized solution for interpreting estimation results. PROC CALIS computes the standardized solutions for all models (except for COSAN) automatically. Standard error estimates are also produced for standardized solutions so that you can examine the statistical significance of the standardized estimates too.
However, equality or linear constraints on parameters are almost always set on the unstandardized variables. These parameter constraints are not preserved when the estimation solution is standardized. This would add difficulties in interpreting standardized estimates when your model is defined meaningfully with constraints on the unstandardized variables.
A general recommendation is to make sure your variables are measured on "comparable" scales (it does not necessarily mean that they are mean- and variance-standardized) for the analysis. But what makes different kinds of variables "comparable" is an ongoing philosophical issue.
Some researchers might totally abandon the concept of standardized solutions in structural equation modeling. If you prefer to turn off the standardized solutions in PROC CALIS, you can use the NOSTAND option in the PROC CALIS statement.