Auto meshing doesn't ensure adequate connectivity between member to member and is therefore recommended for horizontal area elements enclosed by line elements only (floors), where there is no structural connection between floor and any element in between the panel.
In case of vertical elements, connectivity between vertical and horizontal elements is of due importance and is better to be achieved through manual meshing.
In case of auto meshing as you indicated, change in size of auto mesh could solve the problem as its possible that connections inadequate (nodes not coinciding) on 1m element size can be adequate for 1.2m size (nodes start coinciding) and so on.
For p-delta,a possibility exists that due to any meshing error some connection is modelled with inadequate lateral stiffness i.e when program try to impose lateral deflection due to seismic loads,modes start yielding frequency below shift.
If you are interested in studying the problem,then easy approach is to check each mode shape and investigate the member that is going in unrealistically large displacement. Solving this member's connectivity inadequacy will solve your problem.
I dont know if they are called integration nodes, but sure they are called "Gauss Points". these are the points in FEA where stresses are or displacements are calculated and are then interpolated or extrapolated to nodes or within the areas.
Stress at nodes will be extrapolated from Gauss points. But at nodes, there will be many values different values from each corner of adjacent area. You should not average results at this intersection which will help you understanding the quality of mesh.
Its an art where and what to check in FEA. You should consult FEA book by Robert D. Cook. Its very easy and well written plus the author is well known in this field.
I am here Rana Waseem...passed out from NUST , MCE in 2009...working currently in middle east, SAUDI ARABIA! Love structural engineering...own a personal blog too about structural engineering that is ranawaseem.com
