By default, the chart renderer will use a summation method to condense entries within a time range. For this data set, you are probably looking for one of the other math modules, such as mean.
To illustrate, let's consider a user who is monitoring infrastructure performance. An example query might be looking to chart the free memory available on a system:
tag=mem regex "Mem:\s*(?P<total>\S*)\s* (?P<used>\S*)\s*(?P<free>\S*)\s*(?P<shared>\S*)\s*(?P<buffers>\S*)\s*(?P<available>\S*)" | chart free
The chart appears incorrect because the charting renderer adds the "free" value of each entry to provide a total for a given time slice for display on a chart. We can alter the query slightly to insert a mean module in order to take the average value within that same given time slice. This should provide the results we are looking for:
tag=mem regex "Mem:\s*(?P<total>\S*)\s* (?P<used>\S*)\s*(?P<free>\S*)\s*(?P<shared>\S*)\s*(?P<buffers>\S*)\s*(?P<available>\S*)" | mean free | chart mean