It depends what you view a 'step loss' as representing. irl unit firepower is not proportional to losses – heavy weapons are conserved at the expense of riflemen, subunits are kept in reserve, most weapons operate at a fraction of their theoretical capabilities etc.
There are various models around, at one extreme you have that used in e.g. Memoir 44 or AHGCs 'Russian Front' where units retain full firepower until they reach their step loss maximum – the justification in RF is that most firepower comes from the div and corps artillery and that infantry/tank losses just make the unit more brittle.
At the other extreme you have the proportional loss/effectiveness model e.g. Megablitz or Victory Games 'Hells Highway'.
An interesting approach is taken int he 1956 British Army Tactical wargame, where the 1st 'hit' has no effect on combat power at all, the second reduces it to half effect and the third renders it combat ineffective (at losses of around 50%).
I find a two or three step approach works quite well, if you conceptualise the first 'hit' as effectively meaning the unit commander has to commit his reserve, and it is only on the second hit that frontline combat strength is unduly affected. so e.g. in both NATO Brigade Commander and Drumfire (company and battalion stands respectively), the first hit has no effect on combat strength as conceptually it is a platoon/company being destroyed, and the reserve moves up to take its place. On the second hit the unit is removed as it is then reduced to platoon/company strength and no longer combat effective.
You may introduce the same sort of mechanism as in Command Decision and allow such 'eliminated' stands to be reorganised by combining them with another of the same type to produce a single operational stand.
In other games I just take whole battalions off at a time, effectively using them as strength point markers for brigades. There may or may not be some sort of unit recovery mechanism, depends on the depth of simulation you want.
If I'm running a massive participation game(like my Cambrai game with eleven divisions deployed and 14 players), I don't want lots of depth in the simulation.