There are at least four reasons to add a more general ID to an observation.
First, suppose an identifier deletes their account. Hard to believe, but people do become disgruntled and leave this site in a huff. In doing so, they take all of their IDs with them. Thus adding a more general ID at least solidifies the observation at that general taxonomic rank even if the specifics are not known to the IDer.
Second, suppose an identifier does not follow the maxim of “means well” and becomes a muckracker. I’ve mostly seen this with student groups where people will ID to a different taxonomic rank so that the disagreement kicks the observation way back to a general taxonomic rank that is less likely to be reviewed. Adding a general ID will at least hold the observation at that general rank until the specifics can be sorted out by more IDs.
Third, suppose the observation isn’t actually of quality to be able to ID to the specific rank proposed. A more general rank might be warranted, not because the IDer thinks the ID is wrong, but because the evidence doesn’t support the ID. Keep in mind, the observer often has evidence for an ID that isn’t present in the observation directly…maybe they were seeing this taxon regularly, maybe they have experience in the area, maybe they say it run away and the movement style is indicative of one taxon, or any number of reasons. These are all circumstantial, and as an IDer, I cannot rely on circumstantial evidence to make an ID, but rather am limited to the evidence provided in the observation itself. If that evidence is limited or suspect, I may not want to hard disagree, but rather denote that the evidence doesn’t warrant the more specific ID provided, hence a more general ID is appropriate.
Fourth, (and related to the third) suppose a researcher wants to use observations for a project. For many taxa, I am able to say with confidence that if I IDd something to a particular rank it is highly likely that rank, but perhaps not more specific. That is, downloading observations with IDs that a particular trusted IDer made can allow a researcher a shortcut as discussed recently in forum here by @twtaggart. Now I can exclude observations with more specific IDs that this trusted IDer didn’t make to that specific rank as suspect without having to go back and review/confirm each observation myself.
In short, more IDs is generally a good thing, not a bad thing.