Since the data is unique to a time & place and irreplaceable, it would completely destroy the reproducibility aspect of the scientific process.
This gets tricky in some fields, however. I work in a field where generating the data is a notoriously difficult and haphazard process, subject to many non-experimental variables, such that the use of a different pipette or stock solution can make the difference - or even just the speed of the researcher's manual labor. Temperature and humidity play a role too, and these are not as precisely calibrated as one might like. So if an experiment is performed at 8pm on a Saturday night by a grad student in Colorado, there is no guarantee that a postdoc in Singapore will be able to do the same thing based on reading the paper. (Actually, from past experience, there's no guarantee that the original experimenter will be able to reproduce it either!) But the data may be just as good, and they're difficult to fake, and they're deposited in a public database. Since everyone in the field is accustomed to the complexities of the process and we have decent archival policies, this problem is accepted as a fact of life.
I am quite certain that some of my (published) results from grad school would be difficult at best to reproduce exactly. I stand by my data - and am happy to share them - but it is always troubling to wonder if someone else in a different environment would have reached different conclusions.