A collection of raw data is full of systematic errors, accidental mistakes, misleading black swans, and false trails (some of which get followed for years before they finally turn out to be false). I've seen several talented, well-trained, and highly experienced scientists fool themselves for decades with their own raw data.
I've done a lot of data cleaning over the years, some of it geological. Yeah, there are usually some problems. But I've never had one that I couldn't resolve. Your post implicitly assumes that only one or the other can be published. Not true. As a condition of receiving grant money, document and publish ALL raw data and any cleaned data, in addition. The interwebs still has a few bits left to hold the extra.
I've done a lot of data cleaning over the years, some of it geological. Yeah, there are usually some problems. But I've never had one that I couldn't resolve. Your post implicitly assumes that only one or the other can be published. Not true. As a condition of receiving grant money, document and publish ALL raw data and any cleaned data, in addition. The interwebs still has a few bits left to hold the extra.