Yes - it looks like a capacitive voltage divider - however, it isn`t one because the signal input is NOT at the top of C1 but between both caps). If you try to look into the feedback circuit (starting at the emitter node which has a dynamic output small-signal resistance r,out) we face a third order circuit. This circuit is a first order grounded RC block (r,out-C2) that is connected with a second order grounded CL block (C1-L1). Both blocks together provide something like a bandpass response, however, with a squared s in the numerator and a third order denominator (EDIT: First order RC lowpass in series with a second order highpass).
(That means: Your second diagram is not correct: Vout is not across C2)
This bandpass provides - as desired - a zero deg phase for the resonant frequency wo. The necessary loop gain (>1) at this frequency is possible because of the existing voltage enlargement across the inductor at w=wo (although the common collector stage has a gain below unity). C3 acts simply as a coupling capacitor and does not influence the oscillation condition).
A good description of the circuit (including derivation of formulas) can be found here: