I am not a Jedi.
But, if your original data is undersampled, then deconvolution doesn't work well regardless of drizzle.
On the other hand, if your original data is at least critically sampled or oversampled, then deconvolution seems to work OK even on a drizzled version.