We analyze the low-redshift (z~0.2) circumgalactic medium by comparing absorption-line data from the COS-Halos Survey to absorption around a matched galaxy sample from two cosmological hydrodynamic simulations. The models include different prescriptions for galactic outflows, namely hybrid energy/momentum driven wind (ezw), and constant winds (cw). We extract for comparison direct observables including equivalent widths, covering factors, ion ratios, and kinematics. Both wind models are generally in good agreement with these observations for HI and certain low ionization metal lines, but show poorer agreement with higher ionization metal lines including SiIII and OVI that are well-observed by COS-Halos. These discrepancies suggest that both wind models predict too much cool, metal-enriched gas and not enough hot gas, and/or that the metals are not sufficiently well-mixed. This may reflect our model assumption of ejecting outflows as cool and unmixing gas. Our ezw simulation includes a heuristic prescription to quench massive galaxies by super-heating its ISM gas, which we show yields sufficient low ionisation absorption to be broadly consistent with observations, but also substantial OVI absorption that is inconsistent with data, suggesting that gas around quenched galaxies in the real Universe does not cool. At impact parameters of <50 kpc, recycling winds dominate the absorption of low ions and even HI, while OVI almost always arises from metals ejected longer than 1 Gyr ago. The similarity between the wind models is surprising, since we show that they differ substantially in their predicted amount and phase distribution of halo gas. We show that this similarity owes mainly to our comparison here at fixed stellar mass rather than at fixed halo mass in our previous works, which suggests that CGM properties are more closely tied to the stellar mass of galaxies rather than halo mass.