confidence levels on 2d \chi^2 grid, where the minimum \chi^2 is not near 1.

Hallo everybody!

I have the following problem:

I obtained a 2D grid of reduced \Chi^2 fits, where the the minimum \Chi^2 is far from unit (1).
This is due to some noise in the data and the model it self is not perfect.

Now if the minimum \Chi^2 was ~ 1 then I can apply the delta \Chi^2 confidence levels adding 1 sigma = \chi^2 + 2.3/DOF, 2 sigma = \chi^2 + 6.2/DOF, \chi^2 + 11.8/DOF, where DOF is the degrees of freedom in the model and the coefficients are the \delta \chi^2 levels for a grid with 2 free parameters.

But how this will work if the minimum \Chi^ is about 6 then? Is there are another way to obtain the confidence levels if the model is not perfect?

If I apply the \delta Chi^2 method I am getting very very small confidence regions...

I hope you can open my eyes!
All the best!