I agree with the effect you report but not the reasoning. If one is confronted with a flock of mixed Crossbills and want a photo of Parrot Crossbill, then both for personal aesthetic and to avoid later dispute the selected individual is going to be the classic. This is not least because any birder is going to look at intermediates and go: "hm, not sure about this one, lets do the one with the really big bill that I am sure about". Their own certainty is an issue in selection.
This counts double or treble for winter Redpoll flocks when Arctics have been reported. A crowd of birders will agree almost exactly on the number of birds in the flock but have different opinions about how many Lesser/Common/Arctic Redpolls are present.
We agree that the bias exists, so the reason is maybe less important. What you describe is partially expectation bias: birders expect two separate species, so report only small and big birds.
I agree that Redpolls are similar situation: genetically identical species split into 3. And biases are similar: cherry picking publications proposing the split and ignoring ones against, waiting for an (unlikely) publication openly proposing the lumping etc.
BTW, anbody interested in bird taxonomy could be interested in thought fallacies and biases. Wikipedia gives an overview, although it sometimes falsely narrows biases to e.g. only medical or legal topics.
Selection bias - Wikipedia - picking publications for the split, and ignoring or diminishing papers against the split. Within a single new research like a genomic study: picking new differences, but ignoring new similarities.
Observer-expectancy effect - Wikipedia - tendency of birders to show discrete forms or characteristics (e.g. discrete sonogram shapes) if they expect different species
Reporting bias - Wikipedia - birders are not likely to report unidentifiable or intermediate birds.
Publication bias - Wikipedia and
Funding bias - Wikipedia - publication and funding are related to / variation of the reporting bias found in published papers: research proposing big differences is more likely to be funded, submitted and published than negative results,
Conflict of interest - Wikipedia - the result of the previous ones: researchers are rewarded with grants for discoveries like new species and publications. This can result in selection among the pool of researchers leaving only researchers proposing splits. Note that it is not necessary that even a single researcher, and not even a tiny bit unethically, changes his mind, or is influenced by grants. Simply only the ones which produce flashy papers remain.
Status quo bias - Wikipedia - waiting for a publication to suggest lumping, even if it is not likely due to the publication and funding bias.
Fallacies of definition - Wikipedia - changing what definition of a species to use. Here, the lack of genetic differences within crossbills and redpolls means single species under both classical biological and population species concepts.