I'm also seeing a massive discrepancy between the data from this API call and what Dashboard shows.
For a network that contains only an MX84 HA pair, with dual WAN links:
For 20th June (from 000000 to 23599 UTC), with 600 second sample period, API data says...
Total Tx bytes 1,484,364,604
Tota Rx bytes 982,713,386
Changing to a 3600 seconds sample period gives only slightly different values 1,453,143,728 and 953,523,815 so it isn't especially sensitive to sample period.
Also, looking at the numbers returned, the reported bytes transferred (both tx and rx) in the 600 seconds often far exceeds the capacity of the WAN link.
Dashboard summary report for that network on the same day...
Total data uploaded 490.7 MB
Total data downloaded 404.3 MB
I also checked on an MX65W, the API reports c. 150MB for both send and received, the Dashboard summary reports 0 kilobytes for both, again a big discrepancy.
The API call is reporting more than twice the Dashboard figure, and also impossibly high transfer rates for the sample time period.
Seems clear the API call is returning bad data.