A conversion event pointed at the wrong action is not a reporting problem. Reporting is the last place it shows up, and by then three other things have already been built on top of it.
How it travels
Automated bidding optimises towards whatever you have told it a conversion is. If the event fires on a page view rather than a form submission, the algorithm does its job perfectly and finds people who view pages. Budget then follows performance, so spend moves towards the campaigns generating the most of the wrong thing. Finally the report says the campaign is working, because by its own definition it is.
Nothing in that chain looks broken from the inside. Every step is behaving correctly given its input. That is what makes it hard to catch.
The usual suspects
Events that fire on page load instead of on the action. Thank-you pages counted every time they are refreshed. A single conversion defined twice under different names, so it counts twice. Test traffic that was never excluded. None of these are exotic, and all of them are common in accounts that have been running for a while and changed hands once or twice.
Verify before you scale, not after
Scaling spend on top of a broken signal multiplies the error rather than the result. Fix measurement first, then structure, then budget. That order is the whole approach on the Google Ads management side, and it is not a coincidence that the tracking work comes first.
A number that disagrees with another number is uncomfortable but survivable. A number that quietly agrees with itself while being wrong is the expensive one. The setup and audit work for this is on the GA4 and analytics page.