the problem lies not in the definition of the fictitious clearance, but in the contact type. You are using a penalty contact, which always has penetration of the contact surfaces. A higher penalty coefficient will decrease the inter-penetration, but it will never vanish.
In your first example you have a run called 1srhighpenalty, where I assume you increased the penalty coefficient to see if it has an influence on the results, Unfortunately you forgot to change the value for the contact Bolt-Plate, which has only a value of 1e11, which is not much compared to the stiffness of the involved materials (steel E=1e11).
The penalty value should be about 5-50 times as high as the materials Young’s Modulus.
Increasing the penalty coefficient should improve your results drastically. On the other hand the convergence might become harder and some high value it won’t converge at all.
To check the contact clearance/penetration you can add the contact results global field in the global fields to your run. It has a component called “JEU”, which is the contact clearance on the slave side.
Finally you could try the augmented Lagrange contact type. It does not have contact penetration at all, but is hared to get converged, especially if friction is activated.