for example, some people think lemmings saved should be the highest priority, with skills used as second, and finally time as least significant
Yes, and with two good reasons.
If a scoring system doesn't match the user's valuation, it will be ignored.
When ranking lemmings first, then skills, then time, it is easy to compare solutions at a glance before they are even executed. If a skill were worth 10 seconds (the necessary choice of arbitrary number feels already bad), it would not be so clear. If a skill were worth some arbitrary, very large number of seconds, you get an obfuscated version of (lemmings first, then skills, then time).
The only benefit I see in a mashed-together scoring system is slight ease of implementation, because the programmer has to save a single value only. I don't see any other use of mashing everything into a single number, because the player doesn't think like this at all. Slight ease of implementation should be shunned in comparison to a more natural-feeling system.
Also, ime limits are not the norm anymore in contemporary level design, except where it's the only way to fix backroutes. Compute solution times as if there were no time limit, counting up from zero.
in other words, the score is only updated when a lemming is saved, not every time a second passes. This is so that it isn't a rush to exit the level as quickly as possible after saving the last lemming.
This is a very important and excellent suggestion.
-- Simon