I'm new in FEA and now using STAAD to check steel deck plates. I believe I have to compare the value of Von Mis against yield stress divided by the appropriate material factor (Ref Code: NS3472). But I'm getting larger values of Von Mis when I make smaller mesh of the plate model. So that makes it wrong to compare the Von Mis to yield stress. How should I do it to make a proper comparison?

It's probably still a proper comparison.  However, FEM results are notorious for having localized stress risers around points of discontinuity.  By that I mean applied joint loads or a defined boundary conditions.  

Look at it this way: If you apply a Joint load to the middle of a simply supported plate then what is the stress at the location of the point load?  Since a joint is an infinitesimally small area, the stress at that location would be infinity.  

The reality is that the point load is not applied over an infinitesimally small area.  So, it takes some engineering judgement to interpret the results of an FEM model near these points of discontinuity.   

Well put Josh.
This is a common issue in all sorts of shell finite element analysis situations, and I usually recommend averaging or integrating the stresses over some nominal width for comparison to any code defined allowable stress level, regardless of the material.