if you use every time the constraints, the moemnts obtained on the "floors" (shell) will be much smaller, and therefor you'll obtain much smaller reinforcemnt area necessary to the floor.
In most situations, the only constraint being assigned to a floor(modeled with shell area elements)is a rigid diaphram type, which acts on the in-plane forces only, not on the moments which would be used to calculate slab reinforcement. You're correct in that some constraints can potentially cause other problems that the engineer needs to be aware of. For example, if you use rigid diaphram constraints on floors, you can't check local axial forces in beams, or accurately check floor reactions in irregular shaped buildings either. Rigid diaphrams may also give you less accurate fundamental period, particularly with irregular structures. I agree that in some cases, particularly floor systems and walls, the engineer needs to consider the option to not use constraints such as rigid diaphrams, but instead use actual FE element stiffnesses based on material and section properties(aka 'flexible diaphram'). It's the non-centerline connections where I think joint constraints/frame insertion points have the most application and benefit. If beam joint connections are modeled more realistically using insertion points or constraints, that may affect calculated reinforcement, but it should be more accurate and realistic.
Consider these situations involving the possible application of joint body constraints or frame insertion point assignments: A thick floor slab modelled with shells with deep rectangular or Tee beams modeled as frame elements - If you model beams and floor all connected at centerline joints instead of using frame insertion points and/or constraints to "move" the beams below the slab centerline to where they are in reality. In many cases, this will affect calculated moments in the beams significantly.. not uncommon to change the amount of required calculated reinforcement in the beams by 15% or more. It may also result in your floor slab being more flexible since beams are no longer modeled on the slab centerline. But in such cases, use of insertion point frame assignments or constraints gives you more accurate reactions which can affect the design. Of course, it takes a little longer to model with constraints/frame insertion points, but it's still much faster than modeling these effects with with "dummy" rigid frames (dummy frames can look bad graphically too) or 2-point rigid links. Again, an engineering judgement call as to what is accurate enough.
Another situation which was discussed on another thread is in 'stepped' column configurations where you have concrete columns of different sizes (larger columns at bottom, smaller ones at upper stories) lining up at a common face of the building. In reality, this results in moments being cranked into the structure as a result of these non-centerline column alignments. How significant are those moments and to what extent are they resisted by the floor systems? It depends on the structure. But in order to accurately consider those effects, you need to use frame insertion assignments to 'move' I and J ends of the selected columns, in which SAP will automatically assign joint constraints internally.
This is a good discussion on use of constraints. Too bad it's in a thread entitled "Shells".
