A purist might argue that if you are going to publish any data from an experiment, you should publish all the data, regardless of whether they fit the expected trend. Practical limitations on journal space, however, require that this principle not be taken to the extreme. Among the issues to consider are:

• Was there an error in following the experimental protocol when the outlying data were collected?
  
  
• Would publication of the outlying data require a significant amount of additional journal space?
  
  
• Are the outlying data so irrelevant that to publish them would be a distraction from the focus of the paper?
  
  
• Would readers benefit by knowing about the outlying data and why they are outlying?
  

In this particular case, it seems unlikely that publishing one extra set of data and including a paragraph to discuss them would unduly extend the paper or cause it to lose its focus. At the same time, readers may find it useful to know that the growth trends might be different in different types of soil. Unless the selection of the tenth site did not follow the protocol established for the experiment prior to the start of data-collecting, the data should be published. Alternatively, the authors could at least indicate (with explanation) in their paper that data associated with rocky soil had been excluded from the analysis.