Yes. By reducing the distance between the thing you want to simulate and the box boundary you would get more interactions/collisions which translates to a higher concentration. However, if you model a part of bigger structure one should take care that bits that would not get in contact in nature do not get in contact in the simulation as well. This means that the distance to the bounding box should be large enough so that no interaction is taking place across the box. Hope that helped.

Yes but do check the protein is in good fitting inside the simulation box.... too much small box will effect the flexibility of the protein and may lead to false results...

Regards

Yes and no. There are minor differences which can cause your results to be false. Reducing the grid box would definitely decrease the total volume so that the interactions will increase just like in a more concentrated situation. However, when your ligand and/or receptor (target molecule) goes outside the walls of grid box, it comes back from the other side (Leaving the grid from right, entering from left etc...). So in that sense, you create a new collision probability for the atoms to interact. Let's assume you have a tiny drug that you want to bind to your large protein to inhibit/suppress its activity. In terms of theoretical chemistry, it may not have any chance to attack any residue on the right side of that protein. As you limit the grid-box, some right parts of that protein will enter from the left side and your drug will be able to interact that side as well which is not supposed to happen. So if you want to increase the concentration go ahead and do it. Please do not decrease the grid-box volume to satisfy that need since it will yield wrong outcomes.