If you are seeing a few regions of identity or half-identity between two unrelated people, that does not necessarily mean that the people are recently related.
It is possible that the relationship is very distant in time—by chance, the two people happened to inherit the same piece of DNA from a long-gone ancestor. If two people belong to the same population and the population is relatively homogeneous genetically, then the chances of them sharing some regions of DNA increase. For example, two people of Ashkenazi Jewish ancestry may often have regions of common ancestry, because many "unrelated" Ashkenazi Jews are actually distant (fifth or sixth) cousins, sharing multiple common ancestors many generations ago.
Finally, our algorithms provide only estimates, and sometimes will mistakenly call a non-identical region identical. This is more likely to happen when both people come from genetically unusual and/or homogeneous populations. As scientific knowledge improves and our database grows we will be able to provide increasingly precise and accurate estimates.