You can use partitioning (by date) on an Uncached FTP dataset: you specify just / as the path, the in the Partitioning tab, activate partitioning, add a time dimension and set the partitioning pattern to dataset_%Y-%M-%D.txt.
You can then create a Sync recipe from this dataset to another dataset on a local storage connection (e.g. File System) with the same partitioning (the default is to keep the same partitioning and and equals partition dependency.
Finally, you can build this dataset (or other downstream datasets with the same partitioning) from a Scenario: just add a build step and specify CURRENT_DAY in the partition (note: you have other placeholders, like PREVIOUS_DAY, should you need them).
Hope this helps.