Co-variance curves connecting independent variables are often omitted from path diagrams as they tend to clause a lot of clutter, but they should always be included in the statistical model. Mplus will do this automatically and AMOS has an option that will add all the covariances rather than having to do this manually.
Thanks a lot Mark. I missed little further info. I really needed to improve the indices since I ended up with 0.8.. for most of them. If covariance drawn, indices seem to be improving. Is it O.K.
Yes, the covariances should always be included. Indeed, there are occasions when the residuals among endogenous variables should also be correlated also (e.g. repeated observations, multiple mediators) and model fit may not be good unless these are included.