Interesting measurements! But not so surprising finding, when you remember that variable contrast is achived by superposing (adding) the densities of 2 (3 for some papers) layers: one soft, the other hard contrast; at the highest print densities, the hard layer dominates the response. I had assumed that paper manufacturers had somehow addressed that issue; it does not seem to be the case, unfortunately.

This said:
(1) presumably you placed the step wedge in the negative carrier, not lying on the paper; since you see different responses for different heads, I assume that is what you did;
(2) if you want to address the response curve of the paper, and isolate it from the response of the system (projection system + paper) you should plot the paper density versus the (log)illumination measured on the enlarger board, on the projected image of the step wedge. Otherwise, you are measuring a combination of the paper response and of the effective density of the step wedge, for which the Calllier effect depends on the illuminant. Alternatively, you might place the step wedge in contact with the paper; that would eliminate the role of the head (condenser versus diffusion) but would still not guarantee that the effective density (seen by the paper) is the same as measured by the densitometer.