MWCO is a single-point description of rejection (the 90% level of the curve), while full characterization requires at least two parameters - e.g. the slope of the curve around its inflection point is needed besides that one value where MWCO is. So in fact very different membranes can have similar MWCO.

 Besides, there could be specific interaction between the protein and one (or both) of the membranes. Contact angle would not be able to distinguish these differences.

You have partially answered your own question by using the word 'SIMILAR' to describe the protein, you did not use the word, "IDENTICAL' protein. Additionally, as explained above, MWCO values are a gauge and not the final determinant. There is no substitute for lab research to determine the optimum choice for your protein. 

Same MWCO membranes can have very different solute rejection properties. The separation characteristics of the membranes are not fully dependent on the MWCO. In case of membrane separation, the pore size, pore size distributions and the avarage porosity might be playing an important role. I have had the experience of using same MWCO polymers for gas separations and the results were pretty different. The main reason was the average pore size of the two membranes. Furthermore, only the contact angle is not enough to predict the surface hydrophilicity. The two membranes, having same contact angles but different surface roughness parameter will show quite different surface hydrophilicty. The relationship between surface hydrophilicity (solid-liquid interfacial free energy) and the contact angle is given by Young-Dupre equation modified for surface roughness parameter.

does pore size distribution which differ one another can cause (relate to) different inflection point between them?

> does pore size distribution which differ one another 

 This certainly is the key issue, which is often overlooked, alas.

Just to add one more caution to this: the MWCO is not really selecting by molecular weight, but rather by size. The size of proteins can be estimated from their molecular weight, if the protein behaves like an average globular protein (see table in link below). However, if the protein structure or conformation (oligomerization, unfolding) is different, that will lead to an effective molecular weight cutoff that may be different from the one for a globular protein. Especially when mixing inorganic components with proteins, these changes may be quite significant.

http://www.materials-talks.com/blog/2016/01/21/zetasizer-sensitivity-for-protein-dls/