Dynamic Networks have been introduced in the literature to solve multi-sensor fusion problems for navigation and mapping. They have been shown to outperform conventional methods in challenging scenarios, such as corridor mapping or self-calibration. In this work we investigate the problem of how raw inertial readings can be fused with GNSS position observations in Dynamic Networks (DN) with the goal of i) limiting the number of unknowns in the estimation problem and ii) improving the conditioning of the normal equations arising in least-squares adjustments in the absence of spatial constraints (e.g., image observations). For that we propose a modified version of the well known IMU-preintegration method, accounting for a non-constant gravity model, the Earth rotation and the apparent Coriolis force, and we compare it with the conventional DN formulation in a emulated scenario. This consists of a fixed-wing UAV flying four times over a 2 km long corridor.