High-order harmonic generation efficiency is theoretically modeled and compared with experiments using 400 and 800 nm driver pulses. It is shown that, for a short drive wavelength and a Keldysh parameter larger than 1, the Ammosov-Delone-Krainov (ADK) ionization model does not give a good agreement between theory and experiment. Since the ADK ionization model only accounts for tunnel ionization, it underestimates the yield of low-order harmonics from the wings of the driver pulse. In contrast, the Yudin-Ivanov ionization model [Phys. Rev. A 64, 013409 (2001)], which accounts for both tunnel and multiphoton ionization, gives much better agreement with the experimental results.