I am in doubt about the value of C affecting the regularisation. Doesn’t higher C implies more bias and hence it can allow misclassification of certain samples.