That can be caused by upward drift of the source under test (maybe not so likely with your simulated oscillator!) but is mostly a consequence of the small angle criterion being violated. This criterion is baked into the definition of L(f) phase noise.