depends, if like say one build is 3 seconds and the other build is 1 second, I prefer the 1 second, but if it turns out that the 3 second build is more easily debuggable and understandable then I would gladly sacrifice 2 seconds.

If one build is .75 of a second and the other build is 1.5 seconds I probably wouldn't consider the .75 of a second a significant improvement because I would probably not gain anything from the slight improvement.

I'm sure you can construct lots of other examples whereby one my prefer a less faster build over a faster build or just feel that it was not actually important either way.