This is true, but however sophisticated the algorithm, at the end of the day, they do just "spit out a number". If that number is consistently smaller than both other algorithms, and what practice has shown to be safe, then that number is less useful that one which is consistent with both those things.
As I can't think of a recreational computer that doesn't use a Haldaneian or neoHaldaneian model at its base, I feel that's as reasonable a way as any to make such a decision when deciding on which one to use.
You can increase the conservatism of a liberal dive computer to meet your needs, however, you can do nothing to make a conservative computer more liberal.
With the earlier recreational computers that use a so called RGBM model this was true, but some of the newer models offer the option of going to 50% RGBM so it's not always the case these days.