It depends on your assumed failure mode. There are various failure theories.
A popular on (I believe used by the ASME Code) is Tresca or maximum shear stress theory of failure. It assumes that the material is highly ductile, which 304 is. In such materials, the failure stress in pure shear is half the tensile stress failure. I don't think those theories deal with yield failures, so I'm unclear if the same ratios apply.
Other theories include maximum principal stress failure (usually brittle materials) and Von Mises (constant energy of distortion failure theory, I think).
The failure envelopes, though don't follow precise patterns. There is statistical scatter.
I think the reason you're finding scatter is that materials don't behave ideally, don't fail at one discrete stress, and may not be anisotropic (meaning their properties aren't alwasy the same in all three coordinate directions).