Degree or order of the splines? I mostly work in TurboCAD, which has provision to support curves up to degree 9, the higher the degree the more tension to the curve as I understand it, also as I understand it, each degree is essentially a factor in the polynomial function of the curve. I've also read that Rhino has the deepest curve support available commercially, up to 32 degrees, while IGES, for example, is limited to three or four-degree curves and surfaces. I've read also that higher degree curves can be closely matched by lower order/degree "piecewise approximation", while the US's Sandia National Laboratories produced some research that claimed that any curve in the presumed multiverse can be duplicated with a maximum 54 degree curve. Whaddya reckon? I think I've gathered more than enough hearsay over the years to be dangerous...
Check out Marcus Bole's PolyCAD app for some really good curve and surface handling and analysis. Registration needed but free, windows only.

You might also be seeing translation accuracy differences. Meaning, I assume you imported one curve and then created the other native curve through the end points and internal nodes?

What transfer format did you use? VCP doesn't do native CATIA, does it? Shark does V4 & V5, but I don't know that it does anything other than solids. What CATIA are you using? Wanna see if Shark does anything different?

The deviation would not be the points as those are absolute. The curves are the issue. Either the order and/or type differences between the systems and of course the iges translation. This why we have, for years, used point clouds for body design. Translated curves leave too much ambiguity and clutter to deal with.