Learning enhances hippocampal engram cell synaptic connectivity which is crucial for engram reactivation and recall to natural cues. Memory retrieval engages only a subset of the learning-activated ensemble, indicating potential differences in synaptic connectivity signatures of reactivated and non-reactivated cells. We probed these differences in structural synaptic connectivity patterns after recent memory retrieval, 72 h after either neutral Context Exploration (CE) or aversive Contextual Fear Conditioning (CFC). Using a combination of eGRASP (enhanced green fluorescent protein (GFP) reconstitution across synaptic partners) and viral-TRAP (targeted recombination in activated populations) to label CA3 synapses onto CA1 engram cells, we investigated differences in spine density, clusters, and morphology between the reactivated and non-reactivated population of the learning ensemble. In doing so, we developed a pipeline for reconstruction and analysis of dendrites and spines, taking nested data structure into account. Our data demonstrate an interplay between reactivation status, context valence or both factors on the number, distribution, and morphology of CA1 engram cell synapses. Despite a lack of differences in spine density, reactivated engram cells encoding an aversive context were characterised by a higher probability of forming spine clusters and a more dynamic spine type signature compared to their non-reactivated counterparts or engram cells encoding a neutral context. Together, our data indicate that the learning-activated ensemble undergoes different trajectories in structural synaptic connectivity during engram refinement.