The issue in here is that the new algorithm checks the integrity of the mesh. The error you are getting means that an element has a negative or zero jacobian. When this occurs the algorithm does not calculate the stiffness properly and hence it will fail. The old algorithm assumed the same but did not check it.

I am not exactly sure what the error is in your mesh, it is most probably an inverted element, which should be easy to correct if you have access to a program like matlab. If on of the elements is flat, a similar error would occur as well, the latter needs local remeshing to solve the problem in the mesh. Currently we have no module in SCIRun to fix errors in a mesh, it has been on the list for a while.

best,

Jeroen