Share of Preference splits votes for each respondent, so a respondent might be 60% in favor of Product A and 40% in favor of Product B. But, with First Choice this same respondent gives 100% to Product A and 0% to Product B. So, you can see how across people the ratios of shares of votes between products is more extreme under First Choice rather than Share of Preference.
The only way to really know whether more extreme shares or flatter shares are more appropriate for predicting some out-of the-conjoint-exercise choices (such as real world purchases) is to have good data about those out-of-sample choices to use to "tune" the degree of flatness or steepness of the Share of Preference model.
You tune the "Share of Preference" model using the "Exponent" setting in the market simulator. An exponent of 1.0 is the default. Increasing the exponent to 10 or 20 will product results nearly identical to the "First Choice" rule. Setting the exponent to 0.5 or 0.25 will make the shares of preferences even flatter than the default Share of Preference rule, until setting the exponent at 0.0 divides the shares evenly across the products within a simulation scenario.