Wrong Values for historic States

I recently migrated from OH1 to OH2.
But one rule does not work as expected.

It looks like the rule is getting wrong data.

I try to compare a current temperature to the value from two minutes ago.

logInfo("Showering", (Temp_Bath.state as DecimalType) + " " + (Temp_Bath.historicState(now.minusMinutes(2)).state as DecimalType))

As you can see in the logs, the older value is constantly lower than it should be.

2017-05-21 18:17:58.625 [INFO ] [pse.smarthome.model.script.Showering] - 25.7500 23.291666666666667850904559600166976451873779296875
2017-05-21 18:18:29.125 [INFO ] [pse.smarthome.model.script.Showering] - 25.8125 23.333333333333332149095440399833023548126220703125
2017-05-21 18:18:59.518 [INFO ] [pse.smarthome.model.script.Showering] - 25.8125 23.333333333333332149095440399833023548126220703125
2017-05-21 18:19:30.015 [INFO ] [pse.smarthome.model.script.Showering] - 25.8125 23.291666666666667850904559600166976451873779296875
2017-05-21 18:20:01.347 [INFO ] [pse.smarthome.model.script.Showering] - 25.7500 23.291666666666667850904559600166976451873779296875
2017-05-21 18:20:31.251 [INFO ] [pse.smarthome.model.script.Showering] - 25.8125 23.291666666666667850904559600166976451873779296875
2017-05-21 18:21:01.709 [INFO ] [pse.smarthome.model.script.Showering] - 25.8125 23.291666666666667850904559600166976451873779296875
2017-05-21 18:21:32.334 [INFO ] [pse.smarthome.model.script.Showering] - 25.8125 23.291666666666667850904559600166976451873779296875
2017-05-21 18:22:02.716 [INFO ] [pse.smarthome.model.script.Showering] - 25.8125 23.292708333333333570180911920033395290374755859375
2017-05-21 18:22:33.302 [INFO ] [pse.smarthome.model.script.Showering] - 25.8125 23.292708333333333570180911920033395290374755859375

Even the chart seems to be showing wrong values.
http://openhab:8080/chart?items=Temp_Bath&period=W

I am using rrd4j in case that is important.
But I hab no problems with this rule with openhab1.

I copied the rrd files from openhab1 to OH2 to preserve my data.
To make sure this is not the problem I let OH2 recreate the Temp_Bath.rdd from blank.
But the issue still exists.

Any idea what the problem is?

The very long decimal place for the historic Item makes my think that rrd4j is prematurely averaging the values. The chart is certainly showing that the lower values are what is being stored in the database.

The first place I would look is the rrd4j config and see if there is something going on there.