If I use the "log(inhibitor) vs. normalized response -- Variable slope" in Graphpad 5.0 I find an IC50 of 79.55 with a 95% Confidence Intervals of 65.39 to 96.77. Researchgate doesn't allow me to attach the file so I send it to you as a message in RG.
If the values you give for inhibition at the different concentrations are averages it improves the statistics if you type the individual measurement results into the Graphpad file.