I understand that the slope decreases with increasing d due to the inverse power square law, . . . .
If it was an inverse square relationship the count rate \times the distance squared should be a constant independent of distance.
One assumption which is made in the theory for an inverse square relationship is that the particles are not absorbed.
For beta radiation travelling through air this is not true.
Another point is the emitted betas have a range of energies as shown in the graph below produced from readings taken from a scintillation counter.

Those betas with the highest energy have the greatest range,
What type of detector did you use?
With a potentially high count rate when the distance between source and detector is small and a long dead time many beta particle would be missed.
Does the detector that you used count gamma particles which would be produced as the beta particles are slowed down when passing through air?
The data points on your graph for small distance would be very sensitive to the measurement of the distance between the source and the detector especially as you have to square the distance.
How did you locate the positions of the source and the detector?
It is not clear if the count rate was corrected for background radiation.
Finally I think it would have been better if you had presented a table and/or a graph of corrected count rate against distance between the source and the detector.
If you were interested in investigating an inverse square law a graph of corrected count rate against the reciprocal of distance squared might have shown a trend.