The simplest, blunt approach to seeing how much zkillboard is reflective or not of the total loss within the game is to compare the stats zkill provide, with the ESI stats.
That is, zkillboard publishes the number of kills added in the last hour:
At the same time, the ESI provides a cached value of kills by system: https://esi.evetech.net/latest/universe/system_kills/?datasource=tranquility
The 2 will never line up perfectly, which is why it’s a blunt estimate, but just totaling up the number of pod and ship kills (which includes structures) from the ESI and comparing it to the zkill values, gives a rough estimate of how many kills are missing overall (not ganking specifically). The current cached totals for examples:
1333 / 1966 = 68% of total kills
It varies throughout the day and varies because of the different times periods between the live value of zkill and the cached value that CCP provide.
Looking at it multiple times a day, it generally varies from about 50% to 95%.
The only real valid conclusion from that, it that there are a lot of kills overall that aren’t posted to zkillboard.
To get better than that needs validation of the data for specific systems, or ship types. It’s time consuming, but without access to CCP’s raw data, sampling is the only approach we have.
For freighters for example, sitting in Uedama with alts on each gate and then comparing the number of ganks observed to the number that appear on zkillboard provides a much better comparison. Doing that across multiple times zones, multiple days, multiple systems and re-validating it periodically, provides even higher confidence.
Last time I validated the freighter loss for example, >95% of observed highsec freighter losses were posted to zkillboard. So for freighters at least, zkill data is pretty reflective of actual.