Empids hit a number of iNaturalist’s weak points simultaneously.
It’s often impossible to ID Empids up to species just from a picture, but it is often possible to narrow them down to one of two, which is well below genus level. These aren’t formal species complexes, so they don’t show up in iNaturalist’s taxonomy tree. You may not be able to tell if a bird is a Willow or Alder Flycatcher but you can tell that it isn’t Least, Acadian, Yellow-bellied or anything else. Birders (and eBird) can still ID it as a Trail’s Flycatcher (the old name before the species were split) but iNaturalist doesn’t have that option.
Even if you can’t ID the bird visually, you can use context, such as location or time of year. It may be impossible to distinguish Willow or Alder on sight during migration, but they have different breading ranges. During breading season, there’s usually only one option. This is explicitly covered by eBird’s filters, which track likelihood a species being present down to the week. INaturalist’s “seen nearby” seems to lack a temporal component.
Empids can, of course, be IDed by call (in fact it’s mostly a birding by ear game,) but iNaturalist really hasn’t developed its audio capabilities.
There is also the issue of different habitat preferences, which get encoded directly into eBird hotspots, and which iNaturalist has no knowledge of at all. As well as the fact that eBird simply has better reviewers.
There’s no harm in doing so, but I’d never treat iNaturalist as a repository of bird data. The number of sightings on eBird is somewhere between one and two orders of magnitude greater than the number bird sightings on iNaturalist (slightly greater than the total number of research grade observations on iNat,) is not dependent on photographs, and is explicitly designed to be a scientific tool.