Hola Mauro,
the idea in calculating TPs variance in LRS or local East North Up system reference is that its axii are representative RMS TP east, North up, while using RMS in geocentric is not meaningful and using PCS (ie. UTM or other) is not possible as conversion from geocentric to PCS cannot be represented by a 4x4 transformation matrix which needed in the propagation of variance-covariance matrix from internal CS...