I'm thinking you need to run this as a non-linear study. At the least try checking on Large Displacement in your Static Study options and see if it makes a difference.
I looked at the pdf and I see the deformation scale is 107, that is highly exaggerated graphical plot. What do your plots look like when you set the graphical deformation scale to 1 or true scale? Have you checked displacement values to determine if the interference is "real" graphical only?
Hi I'm having a similar problem this seems to happen when I'm doing (or attempting to) quick itterations in sheet metal parts with no penetration on faces and global bonded on. However it seems that when global bonded is suppressed and only the individual contact sets are on that the thing seems to behave and not penetrate. Have to confirm this. To me the contact conditions are 'fighting' the net result is an illogical displacement screen display parts penetrating although the surfaces are no penetration constrained (I'll have to probe the mesh to see what the extent is - no I havent turned to 1:1 scale - but the observed effect is inconsistent across the model for similar faces in contact).
Its all a bit of a pest really (tends to soak up inordinate amounts of time - unbillable with no job progress). I think I'll go back to Roark and Young and hand calcs.
Your plots do not necessarily show penetration.
Assuming there is a gap between the cap and body, in order for it to close the cap would have to move to close the gap. Since the deformation scale amplifies movement (displacement) by a factor of 108 it is very possible that your plot is just showing the movement till contact occurs.
You need to look at the results of the contact elements themselves. That will show whether they are closed or not.