The reason I think it's a problem is that Dendreon had a clerical error on their PIII. The initial p-value reported on the trial was .8 ish and after the clerical error was fixed, the p-value dropped to .52 ish. The FDA never accepted the lower p-value and evaluated the trial based on the higher number.
I saw another trial where there was a mixup, every patient affected by the screwup was censored.