x1 is the standard variation of/in data set/distribution x2; x1 is the quantified (in)accuracy of the results x2 given by some observation or measurement (procedure), calculated from/by standard x3
Technically, there really is only one possible standard deviation, but in practice, it is often approximate via some means x3 (such as assuming that the distribution is Gaussian, even if it actually is only a finite data sample x2).