Open-source intelligence (OSINT) practitioners routinely treat social media profiles, public posts, and visible network connections as ground truth for subject profiling. This epistemological assumption fails systematically against Sophisticated Actors: individuals who deliberately architect their digital footprints as deception surfaces, producing analyst profiles that are not merely incomplete but confidently wrong. This paper presents three novel contributions. First, it formalizes Content Asymmetry as a measurable, information-theoretic deception signal, the Content Asymmetry Index (CAI) quantifying the mutual information gap between permanent and ephemeral content channels, and characterizing the analyst's achievable profile accuracy under asymmetric content partitioning. Second, it applies graph-theoretic attack surface reduction to social identity, modeling the Sophisticated Actor's deliberate pruning of their visible social graph as adversarial edge deletion and analyzing its impact on betweenness centrality, community detection, and influence propagation. Third, it presents an integrated five-phase Counter-OSINT Reconnaissance framework, combining curation detection, cross-surface temporal correlation, second-degree network reconstruction, cross-platform artifact correlation, and multimodal intelligence integration into an iterative, deception-resistant profiling methodology. These contributions are grounded in a longitudinal practitioner-researcher observation study conducted over eight years (February 2018 – February 2026), encompassing systematic analysis of over 100,000 social media profiles across multiple platforms. The study yields eighteen persona engineering techniques, twelve composite behavioral archetypes and six structural platform-exploitation strategies constituting a formal persona engineering taxonomy, a five-phase counter-OSINT methodology with operational tool chains, and a three-tier confidence stratification framework that explicitly calibrates profile element reliability against source independence and collection depth.