ib4t "your sample size is not big enough".
There's a huge thread on this. It's obviously a bug. Neither the community nor the dev team will acknowledge.
Because you're wrong. No offense, but if you really think OP's figures represent a useful sample size, then your understanding of statistics is severely limited. It's been explained many times, in many different ways, in this thread and elsewhere. You absolutely cannot draw any conclusions based on the kind of evidence presented here.
I'll say it again to be perfectly clear: there is nothing wrong with RE rates.