Nearly collinear paths in lavaan

[Eric Burkholder]

I have run into an issue where a mediator and the outcome are so strongly correlated that lavaan is not giving accurate predictions of the relationship. I am already using DWLS because both variables are binary...is there a way to handle this that I am not aware of? 

[Yves Rosseel]

No. In fact, if the mediator and outcome are very highly correlated, it
may be difficult to distinguish between the two, and it will be
difficult (if not impossible) to show that there is a mediation effect
going on.

Yves.