but for now I am including it here.
When defining a standard geometry file you must ensure that the restraints you impose on the model are independent of each other. Usually this is simple. Bond lengths and bond angles are independent of each other in all cases. Since these types of restraints are what people are most concerned with they get lax.
The issue arrises when one introduces torsion angles and planes. For example, consider the configuration of atoms about atom CG of aspartic acid. There are three bond angles - CB-CG-OD1, CB-CG-OD1, and OD1-CG-OD2. In principle these angles could assume widely varying values and can be considered uncorrelated. When you also impose the planarity of this group you have a problem. The sum of the three angles must equal 360 degrees. This relationship allows you to calculate one angle if you know the other two. When restraining these atoms to a plane you must leave one of the angle restraints out. It doesn't really matter which one.
Similar dependences can arise with the inclusion of torsion angles in the library.
The standard TNT library is not very good at handling this problem. In the case of aspartic acid all three bond angles and the plane are defined as restraints. I suppose I should fix this in the next generation library.
Why do I bring this up in a discussion of converting a library to REFI format? Because REFI handles the problem of overdetermination of restraints in a very exact manner. You simply cannot write a REFI definition which contains too many bond angle definitions because the redundant angles have no place in the file. The format implements this restriction by specifying exactly which angles are proper and which are redundant.
The problem is that the author of the library you are using may not have had the same opinion about which angle is the redundant one. The choice is completely arbitrary. In the aspartic acid example, your library may not have a restraint defined for the angle CB-CG-OD2. The REFI format requires this angle and will ignore your definition of the OD1-CG-OD2 angle. The result is a warning that you have failed to define the CB-CG-OD2 angle. This message does not mean that the TNT library was wrong only that differing and incompatible conventions are being used.