I detected a note about this being an Android App in your screenshot (my only language is English, unfortunately), so I think you're heading in the right direction. Instead of using Visits as your metric, try using Launches. Does that make your Crash Rate more accurate?
Also, keep in mind it's possible to have >1 Crash per Launch if a second Launch occurs prior to the lifecycle session timeout. Some helpful information on that here:
Lifecycle Metrics