You cannot compare human (FIDE) Elo ratings with settings of any engine. Besides of the problems of comparing human Elo and any computer ratings (see below), I see another problem:
Making a chess engine as strong as possible is a well defined task, but having it play at a certain level is not so clear. You could cripple it by limiting the depth, or force it to make a blunder every second game or... Lots of options that can affect the playing strength, in the same way that you have all kinds of different human players. For instance Alice at rating 1800 could be a very stable player, drawing most of the games because of her rather conservative playing style. On the other hand you can have Bob at the same rating of 1800 playing attacking chess winning brilliant games, but also blundering many games.
Trying to map these options to a single number (1-100%) is impossible. Personally if I was to play against engines, I'd rather have settings to let me chose the playing style of the engine (e.g. attacking, defensive, passive, "opening guru", endgame nerd,...) than having it play at a certain strength (whatever that means).
In order to find a setting of your engine which you like, why not play a few games at some setting and depending on the outcome change it? But if you want to get better don't set it too low.
Personally I find playing against humans on or off-line much more enjoyable. And if you do this you will not have a problem finding a player that is of comparable strength.
Why computer Elo cannot be compared with human Elo?
This should really go into some kind of FAQ as it pops up very regularly here.
In principle you can give an ELO rating to a chess engine. However ELO (or glicko or any other rating) is made in such a way that it will only give you the relative strength of people or computers in the pool of people/computers that use this same rating system.
So if you have a FIDE Elo as many human chess players do, this will tell you how good you are among all these FIDE Elo human players, because the FIDE Elo is calculated from the outcome of many games between FIDE players only.
There are also people running engines against each other (CCRL, CEGT,...) creating Elo lists for computer engines. These Elo lists will tell you how good computer engines are relative to each other, because the computer ELo are calculated from the outcome of many games between computer engines only. They do not tell you how good computers are relative to FIDE Elo players because they are separate pools.