Referee 1's Review:

This paper is well written and contains a very compelling phylogenetic analysis of both molecular and anatomical data relevant to marsupial evolution. With only a few quibbles, I found the analysis to be very sophisticated and I wish more systematists were this good. If correctly associated, the significance of the first associated skeletal remains of Australia's oldest marsupial is indeed profound and worthy of publication in a high-profile journal like PLoS.

In the words of PeeWee Herman, here is my "big but":
Association by relative abundance. Doesn't the variable presence of the prootic canal in the petrosals argue against their referral to a "single morphotype"? It seems that there are two types represented by the petrosals, some with and some without a prootic canal. In addition, the fact that 25% of all dental specimens are Djarthia means that 75% are not. There is certainly a good case that some of the seven petrosals belong to Djarthia, but at this point it's impossible to know which.

In the case of the tarsal elements, there are three calcanea and one astragalus. What is the size variation among the 75% of teeth from Tingamarra that don't belong to Djarthia? Even if another dental morph is very rare, if it fits in terms of size with these four tarsal elements, the case for association with Djarthia is weakened. The authors should state explicitly something to the effect that "no other dental material fits the size of the tarsals" if this is indeed the case.

If the authors can improve the justification for association by relative abundance, in particular by excluding other dental morphs and explaining why polymorphism for the prootic canal is reasonable, I think this paper could be accepted with minor revisions. Otherwise I remain skeptical of their proposed association.

It's no fault of the authors that there are so few paleo-constraints for australidelphians (indeed with this paper they're improving the Australian Tertiary record considerably). Nevertheless recycling other clock dates as calibrations ("soft" or otherwise) in this paper adds grist to the mill of critics like Graur & Martin 2004 (Trends Genetics). Some of the calibrations are very tortured, such as a mean of 165 Ma for Eutheria-Metatheria that's "congruent" with Ambondro as a eutherian, "however, current evidence does not support the eutherian status of Ambondro". Just delete reference to Ambondro. If the authors want to present clock dates at all (and I'd recommend deleting this part entirely), they might consider using the method of Douzery et al. 2003 (J Molec Evol) and checking the extent to which their hard-minima are consistent with one another.

At least for eutherians, it's not 3rd positions per se but 3rd position transitions that have been characterized as "misleading". Under MP it would be easy to add a step matrix to downweight TS and keep TV, preferable to deleting 3rd positions altogether.
Indels do not seem to have been used as data; why not? This is doable with both MP and MrBayes.

Also, this is a very large molecular matrix which surely presented some ambiguity regarding homology. But the authors say nothing about alignment criteria or whether certain regions were excluded. Some justification and details are necessary in the main text, even if this info is available in more detail elsewhere and/or in suppdata (which I don't think it is).

The MP tree is hard to read; "ladderizing" all of the pectinate, single terminals leading towards Australidelphia would help.

**********
N.B. These are the comments made by the referee when reviewing an earlier version of this paper. Prior to publication the manuscript has been revised in light of these comments and to address other editorial requirements.