DSS tries hard to provide an interactive user experience and fast responses. One of the main things to do so is to work on a sample when exploring / visualizing a dataset. Also, when datasets get big, they could make your browser run out of RAM, so sampling is pretty useful.
That being said, DSS does not enforces a strict limit, it just chooses a reasonable default. So for not so big datasets, you can parameter the sample to be the whole dataset.
Tip: when the dataset is stored on a powerful SQL server, for visualization, you can choose to use the whole dataset as sample. DSS will then offer to run the aggregations on the SQL server instead of on the DSS server.