starBEAST2: 'Mean in Real Space' prior causing wrong dates?

[Justin]

Hello BEAST users,

I have been trying to run a divergence dating analysis on a familial-level tree (~35 species) with one mitochondrial and one nuclear gene (preliminary testing right now) under a birth-death process with BEASTModel test. Because no fossils exist for my group, I am using two divergence dates from a previous study on the group as secondary calibrations (I know this is not the best calibration, but I don't have fossils). 

The two nodes I am using are:
1) The origin of the whole family: 95% HPD = 46.084-49.361 MYA
2) The origin of a clade of three genera (3 species) within this family: 95% HPD = 8.344-9.222 MYA
 

I tried setting the mean value of these monophyletic priors as 'Mean in Real Space as shown 

 

All ESS values are >200, but the tree recovered dates that are less than 1 MYA (it has been consistently supported that the family is 30-50 MYA), and the node bars are offset. 

Am I inputting or setting one of the priors incorrectly for it to be giving me such low dates? When I try to use the 'Offset' option to do all of this, I get lower ESS values but dates that are not so low and normal node bar placement. Also, am I just supposed to toggle the M and S values to get the interval that I want? 

Thank you! 

Best,

Justin