Except that it isn't regional variation, because all the different types/species (whatever you want to call them) of crossbill (including at least three types of Common crossbill) can be found in the same place apparently choosing to pair with mates that give the same call.
Crossbills are a highly specialised group using a quite peculiar foraging strategy, with a lot of unexpected (but more or less direct) consequences.
They feed on a resource that can't be assessed by sight - they have to land on a tree and try the cones to see if they are suitable for them. A single bird will have to try many cones to decide whether a tree is interesting or not; in a flock of birds, the decision is taken much faster because the group integrates the behavior of its component birds, so that each bird has to test fewer cones. This increases the efficiency of foraging, and is thought to be one of the main reasons why crossbills feed in flocks each time it's possible (even while breeding).
Even minute changes in a crossbill's bill structure affect the type of cones that it can feed on efficiently. For their foraging method to work, a given tree should obviously be interesting (or not) in the same way for all the birds in the group, otherwise integrating the assessment over the group would be a very poor strategy; as a consequence, feeding groups must always be homogeneous in terms of bill structure.
Using calls as a rallying signal prevents the merging of a group with other groups having (potentially) different bill structures. Some characteristics of these calls (frequency?) might be constrained by bill size/structure more or less directly (thus be a real signal reflecting bill structure itself), but otherwise the calls are apparently learned (as is the case in most carduelids; in fostering experiments by Jeff Groth in North America, fostered crossbill chicks ended up giving the calls of their foster parents, not those of their biological parents).
Crossbills mate mostly within their feeding group, so there is absolutely nothing surprising in the fact that they mate assortatively in terms of call and bill structure. The opposite would in fact have been most surprising.
Several forms that coexist at a single location cannot be subspecies, because in taxonomy subspecies always refer to geographical variation. Local morphological variation associated to ecological specialisation could mean either that you have species, or that you have ecologically specialised morphotypes, the latter having no standing in the taxonomic model. I believe that whether you have one or the other should depend on the long term isolation and stability of the forms. The fact that the mitochondrial phylogeny is a complete mess is indeed not really a good signal that this has been the case in the past...
L -