So from what I see, your panel is set to a single day (May 17), and your first breakdown is by mobile device (currently filtered to 2 items).
At that top level, using the first column of occurrences, you can see that the total is 66 - split to 53 (80.3% of the total) and 13 (19.7% of the total).
Then you did a second breakdown by Day.. since there is a single day in your report, there is only 1 row of data in that breakdown, resulting in propagating the 13 occurrences as both the breakdown total and the "May 17" value... and being only 1 day, May 17th accounts for 100% of the Day breakdown.
The same applies to the Visits and Unique Visitors columns....
If you are wondering why a second level breakdown provides both a Total row, as well as rows for each column... let's look at a different type of breakdown....
Instead of that second level breakdown being "Day"... let's pretend that we are breaking it down by a Campaign....
Of those 13 Occurrences, maybe only 8 of them got traffic from a campaign... (assuming we were dealing with more than 1 visit and visitor...)
So your data could look like:
| | Occurrences |
| Mobile Device | 13 |
| Apple iPhone | 13 100% |
| Campaigns | 8 |
| campaign-1 | 5 62.5% |
| campaign-2 | 3 25% |
| campaign-3 | 1 12.5% |
The Mobile Device is the total for your first breakdown, and Apple iPhone is the only row, so it's 100% of the traffic (13/13)... the next level breakdown, under the iPhone is Campaigns, and there are three campaigns that drove traffic to a subset of the iPhone traffic... it's total is 8 (only 8 of the occurrences were driven by campaign data), and the three campaigns percentages are calculated against the total campaigns (5/8, 3/8 and 1/8 respectively)
One more final note, I would suggest avoiding the use of the "Occurrences" metric unless you completely understand what you are pulling... Page Views is specific to actual "Page Views"... while Occurrences is the most inclusive metric, including page views and actions (clicks and other triggers)... so if you are trying to get page views, your occurrences will likely be getting you inflated data (depending on your implementation).
