Another technique is to traverse between the known control points:
Traversing is a much more common technique, where you usually set up at one control point, back-sight on another control point (to get proper orientation), measure angles and distance to a new "traverse point", move to the new point, back-sight to the previous control point, measure angles and distance to another new point, and repeat the procedure until you've measured to, either another known control point, or back to your original point. The angular and coordinate misclosure will be an indication of the quality of your traverse observations. Finally, do a traverse adjustment.
In your case, because your control points are not intervisible, you'd need to start at the first control point and initially back-sight onto a distant, well-defined (but arbitrary) target. Then traverse around your site, visiting the other two control points, and back to the first. (The number of intermediate traverse points will depend on your site, but i'm guessing you'll need one or two between each control point.) When you calculate the traverse, you need to first use an arbitrary, assumed bearing for your distant (arbitrary) back-site to get arbitrarily-oriented coordinates for your traverse points, examine the misclosure, adjust the traverse, then rotate[/] your new arbitrarily-oriented coordinates into your original control point system! Discrepancies, if present, will then reveal themselves.
Again, some or all of that calculation ability may be available within your total station. If not, use land surveyor's software, such as Copan, it can do all i describe above.