Remember the 80's American TV show that had this famous line: "Just the facts, Ma'am"?
In science, as in so many other fields [including detective work], data are absolutely paramount. Theory and unifying principles do have a high position in science. If experimental data don't agree with the theory, it is the latter that heads for the dustbin. Data from well-designed experiments are truly priceless.
This is why tinkering with data in any form (fabrication, falsification) is the most serious of crimes in science (somewhat lower down the order would be the other serious crimes: plagiarism, and stealing and claiming credit for others' ideas). Every time such a fraud is uncovered, scientists take it seriously, dissect and discuss it almost endlessly; it leads to much hand-wringing about what measures can help safeguard the scientific enterprise from such frauds in the future. Some measures do get instituted, until a similar crime comes along shaking our faith a few years later.
We seem to have one such moment right now. This NYTimes story reports that Luk Van Parijs, an MIT researcher, has been fired for scientific fraud.