Hi,
This was written about 15 years ago, so I'm not fully sure I remember why it was done this way. I think the point is that if you choose CurvatureRelations -> False to avoid automatic conversions of Riemann into Ricci or Ricci into RicciScalar, etc then we still want to store somewhere what the relations would have been. In this way at some point of the computation you may want to use those relations manually, by calling CurvatureRelations[CD, Riemann] for example. I believe that nothing will break if you eliminate those definitions of CurvatureRelations, though.
I guess then xTras just followed what xTensor was doing, but again I'm not sure.
If I evaluate these then there is no message:
<< xAct`xTensor`
DefManifold[M, 4, {a, b, c, d, e}]
DefCovD[CD[-a], {";", "\[Del]"}, Torsion -> True,CurvatureRelations -> True]
UpValues[CD]
but if we load xTras instead of xTensor then there is the message you mention. I think this is another typesetting leak, similar to those recently pointed out by Thomas. This is a weak point of xAct in general. In cases like this use
UpValues[CD] // InputForm
Results will not be typeset, but you'll be able to analyze the results.
Cheers,
Jose.