It is not uncommon to see retention going up, especially if your selected timeframe includes incomplete data points. The main reason why the retention is going up is that we are dealing with uneven sample sizes for each "day" of retention when the date range is very recent.
For example, User A became a new user on September 30th. This user would be counted in the denominator of the Day 1 retention rate because this user has had the opportunity to be Day 1 retained. However, this user would not be counted in the Day 10 metric because today is October 2nd and the user hasn't had the opportunity to be Day 10 retained yet. User A is completely excluded in the calculation of Day 10 retention.
The line graph will show you the weighted average of all the retention numbers from the user cohorts within the selected time frame. You can see the data for each individual cohort in the table below the chart. Your retention chart is going up because the user cohorts who joined later are given less time than previous cohorts to retain. Their numbers are lower, which results in higher percentages (temporarily).
If you do not wish to see incomplete data points on your chart, I recommend modifying the date picker to include only users who have had the opportunity to be Day 30 retained (example chart here).