n either case, a paramagnetic should not exhibit hysteresis, since the magnetic moments follow the applied field in a spontaneous and reversible manner.
Thus a hysteresis cycle (when observed) reflects the domain formation process and the magnetic walls in ferromagnetic materials and and we wouldn't have to do the paramagnetic correction!
If I understand correct you are asking about a case in which magnetization vs. field data contains at lest two contributions: one paramagnetic and second e.g. ferromagnetic. You are interested in the second component and the paramagnetic signal is some sort of background. Is that the case?
Unfortunately, the answer is not simple and depends what you are measuring. Provide more details about the measurement, to make this issue more clear.
Nevertheless, I will try to answer you question for a simple example.
Let say that you have a thin layer of ferromagnetic compound which shows a hysteresis loop. This thin ferromagnetic layer is on a thick paramagnetic substrate. The magnetization vs. field data will be a superposition of ferromagnetic and paramagnetic signal. However, in this case the paramagnetic signal will the dominant one. The ferromagnetic component at certain field will saturate (of course below Tc). Above this field the only change will be caused by the paramagnetic part. Therefore you use the high field data fit to estimate the paramagnetic component. See the attached figures. In figure 1 you see the RAW data (both components). I know that the ferromagnetic component will saturate above 700 Oe. Because, I am still in the field and temperature conditions in which the paramagnetic signal is linear, I can use a linear fit (for data between 700 Oe and 1000 Oe) to estimate the paramagnetic signal. The obtained slope can be used to subtract the paramagnetic signal. The result of this correction is shown in figure 2. Remember, that this is an estimation of the paramagnetic contribution.