What do 'resnorm' and the 'residual' vector contain for lsqnonlin parameter estimation of Simulink model?

Parameter estimation of a simulink model, using:

[pOpt,Info] = sdo.optimize(optimfcn,p,Options)

with 'lsqnonlin' as optimization option gives an information output with, among other, residual vector and resnorm. I assumed these contained the residuals and the sum of squared error, but the values are way too low. For example a model with this fit calculated a resnorm of only 4.526, althoug there are clearly single residuals that are >100.

The residual vector contains a number of values equal to my observations, and squaring and adding al the values in the residual vector does give the value in resnorm. This all suggest the sum of squared errors, except for the value itself.

What are these values?