The Flow in Dataiku is a "Directed Acyclic Graph" https://en.wikipedia.org/wiki/Directed_acyclic_graph so a recipe can not have the same input and output dataset.
I recommend performing the "update dataset" as a code (Python or SQL) step in a Scenario. In a scenario, you can read/write datasets regardless of their position in the flow.
[ADDITION] Another option would be to not "declare" the dataset as input to your recipe, but only as output.
Hope it helps,