For an optimum hem, you'd have the dress on the actual wearer, all other elements fitting perfectly already.
In that case, you'd expect the hem to be totally level when worn, but it might look slightly uneven on a dummy where the dummy's shape doesn't match the human. These differences are usually relatively small though especially in a young child. Charlotte's hem wasn't even when worn.
On a dummy the unevenness in the hem is really significant and very lopsided, especially at the back. So the only way that would be correct would be if the wearer had a very large arse on one side so needed extra fabric to go over it. No evidence of Charlotte having a giant right buttock, and given it's uneven on her whilst wearing then the wonky hem clearly wasn't to fit her body shape.
If, and it's a big if, Charlotte wasn't available for fitting then they could have used another kid to get the hem level. But then if we assume this was correctly fitted to that kid then they must have had the giant buttock, and surely if you were using a model instead of the wearer you would not use one who required an extra 2" of length to go over their right arse cheek?
Or you use a dummy, which would be totaly symmetrical. This might result in slight unevenness (e.g. if the child had a fuller stomach) but wouldn't be dramatic and would be level on a dummy. But it's not even close to level on a dummy.
Even doing it completely flat (rather than hanging) should have given a better result than this. I think this would be apparent even on a coathanger.
I honestly don't know how they managed to get it so wrong. I've had hems which have been that uneven after pre-hem hanging (circle skirts in heavy wool with bad bias stretch for example) but it's very obvious in that case.