Binding affinity and allostery are not necessary coupled. In other words, the structural/functional determinants for binding and eliciting an allosteric effect do not necessarily coincide. You may have two ligands binding at the same protein site, but only one of them exhibiting an allosteric effect.
The observation of the compound bound in the allosteric site in the crystal structure is not conclusive evidence that the inhibitor actually binds there under the conditions of the enzyme assay. Crystallization conditions usually differ markedly from the conditions of enzyme assays. The protein concentration is much higher, the compound concentration is much higher, and the buffer conditions are totally different. In other words, the crystal structure can give you a false positive result. I have seen this happen.
If the compound does not inhibit the enzyme in the enzyme assay within its limit of solubility, and if the enzyme assay is set up in such a way that it would be able to detect the presence of an inhibitor, then you may conclude either that the crystallography result is a false positive or that the compound binds in the site but does not inhibit the enzyme, as Adrian Velazquez-Campoy said.
The remark about the Km of the substrate being higher than 1 mM lacks context. What Km did you expect?
Inhibition by compounds at the edge of their solubility should be regarded with suspicion. Compounds can cause non-specific inhibition by forming apparently soluble aggregates. This is a well-documented type of inhibition referred to as "promiscuous" inhibition.