As to me I run conservative transient without initial condition, but wait a long time when oscillator start up. It can take a few seconds of REAL time and huge simulation time. You can define initial condition of inductor as butterfish said, but to get the real duty ratio you need a steady state result. To get it I think you need the same long-time simulation.
When we set this PWL as 0-->VDD ramp in ~ 10usec,
this is practically similar to power on situation
and it got to oscillator under .trans. (within 1sec)
If still fail, crystal parameters might need to be doube checked
assuming AC gain is at >= 5dB
and transient operating point is not far away from AC operating point.