Are you treating Fixation 2 as baseline? Since all the event types follow a fixed sequence, the regressors are bound to highly correlate with each other. If you just got a warning (and 3dDeconvolve did generate some results), that does not necessarily mean something terrible. With a design constrained like this, it has limitation on the statistical power of significance testing on those regerssors. But if you are happy with the results, there is nothing wrong about the analysis.
Gang