Comparing the structures with the E. coli reference structure we find that 55 basepairs are predicted correctly in all three structures. After running refold three additional basepairs were predicted right.
Note that since RNase P forms sizable pseudo-knots, a perfect prediction is impossible in this case.
If constrained folding results in a structure that is very different from the consensus, or if the energy from constrained folding is much worse than from unconstrained folding, this may indicate that the sequence in question does not really share a common structure with the rest of the alignment or is misaligned. One should then either remove or re-align that sequence and recompute the consensus structure.
Sven Findeiss 2013-11-22