Obviously in real space, something as flexible and dimensional as one of the c-type lectins, with their wound arms and heavy carbohydrate recognition domains, would fall onto TEM grid in various positions, so finding the “perfect” micrograph of the molecule is not going to happen.  So determining what tethers the N terminal and parts of the collagen domain are quite related to the subjective interpretation of each individual molecule that one sees.  The arc of the arms, the area of the CRDs, the angle of the separation of the individual arms……all just interpretation.  Because it is a matter of judgement, estimates of the number of amino acides in the collagen like domain that are held together in the central (seemingly more rigid) portion of the dodecamers (and fuzzyballs) varies from micrograph to micrograph, and is more apparently different when the SP-D has been “modified” by the loss of some of its length (as in the mini-SP-D developed by White et al (The Journal of Immunology, 2008, 181: 7936–7943).

These particular measurements show something like 30% of the collagen like domain is taken up in the straight-bar part of the center of the dodecamer.  This might be close..

Just visually and with the marked sites on RCSB i would bet these aa are part of the N term contriution to the center line in SP-D. Purple – signal peptide, [MLLFLLSALVLLTQPLGYLEAEMKTYSHRTMPSACTLVMCSSVES] and these are the AA in the collagen like domain which may contribute to the rigidity of the center portion of SP-D dodecamers and fuzzyballs [GLPGRDGRDGREGPRGEKGDPGLPGAAGQAGMPGQAGPVGPKGDNGSVGEPGPK]