One of the angles that is checked by comparing it with the Engh and Huber values (see a previous section) is the backbone N-Cα-C angle that we call τ.

Wouter Touw studied τ extensively, and came to the conclusion that τ is a very complicated angle to deal with...

The abstract of the article that is being refereed by Acta Cryst is:
Synopsis The angle τ (backbone N-Cα-C) is the most contested Engh and Huber refinement target parameter. We show that this parameter is 'correct' as a PDB-wide average, but can be improved by taking into account residue types and secondary structures, and many other aspects of our knowledge of the biophysical relations between residue type and protein structure.

Abstract The Engh and Huber parameters for bond lengths and bond angles have been used uncontested in macromolecular structure refinement from 1991 till very recently, despite that many authors have critically discussed their ubiquitous validity. An extensive analysis of the backbone angle τ (N-Cα-C) illustrates that the Engh and Huber parameters can indeed be improved, and a recent study [Tronrud et al. (2010). Acta Cryst. D66, 834-842], confirms these ideas. The present study of τ, however, shows that improving the Engh and Huber parameters will be considerably more complex than simply making the parameters a function of the backbone φ,ψ angles. Many other aspects, like the cooperativity of hydrogen bonds, the bending of secondary structure elements, and a series of biophysical aspects of the twenty amino acid types will need to be taken into account too. Different sets of Engh and Huber parameters will be needed for conceptually different refinement programs.

If the article gets accepted, one of these days (...), I will place much more of it here.