This paper investigates the integration of affine frequency division multiplexing (AFDM) with faster-than-Nyquist (FTN) signaling, referred to as AFDM-FTN, as a spectrally efficient waveform solution for high-mobility wireless communications. In such scenarios, highly dynamic channels and FTN-induced inter-symbol interference (ISI) pose significant challenges to reliable high-spectral-efficiency transmission. We first establish an analytically tractable upper bound on the capacity of the AFDM-FTN system and further optimize the transmit power allocation (PA) to maximize this upper limit, providing a capacity-oriented benchmark and design guideline. To enable efficient detection under realistic conditions, we then develop an efficient OAMP-assisted receiver tailored to AFDM-FTN in the presence of channel estimation errors. To characterize the detector performance, we derive the associated state evolution recursion and evaluate its mean square error (MSE) behavior and convergence trends. Numerical results validate the effectiveness of the proposed information-theoretic bound and demonstrate the gain of the optimized PA. We also show that the proposed OAMP-assisted detector achieves superior performance compared to conventional linear minimum mean square error (LMMSE) estimators.