it's really not that bad, the model will give you an idea how well it will work, and give you an idea whether the standing currents are within the capabilities of the transistors. you just need to be cautious about the frequency response figures you get, and be aware that even the best models may have some parameters missing or inaccurate, usually because many models are made with more emphasis on some parameters and less emphasis on others. RF transistors for instance will have most of the high frequency parameters very well modeled, but may be less accurate on DC parameters. another caveat is that when you look at a data sheet for a transistor, you will likely see that the beta (current gain) is given as either a minimum value or a range (i.e. "20-80"). variations in beta in a certain part number will usually fall within a wide range, with about 70% of them being somewhere near the value in the "typical" column in a data sheet, so you could see something like this in a data sheet:
___________minimum_________typical__________maximum
beta (Hfe)_____20_____________50_______________80
so the beta would be at least 20, usually around 50, and sometimes as high as 80. if you buy 10 of these transistors selected randomly, you might have 1 or 2 as low as 20, about 6 or 7 of them between 45 and 55, and 1 or 2 of them as high as 80. if they all came from the same lot code, they will usually be within 10% of each other throughout the whole lot code, so you will likely get all 10 of them within the 45-55 range, or within the 65-75 range, etc....
models, on the other hand will most likely be preset at the "typical" value, so all instances of any given transistor in a circuit will all have the same beta. if you wanted to simulate the tolerances in transistors, you would want modified models that either randomize within the beta range, or just make a couple of copies of the "typical" model and create a min and max version and select which one of the 3 models to use at random. resistors, caps. inductors also have tolerances that can be "fudged" at random. it's best to get a circuit working properly with the "ideal" or "typical" values first, then begin "fudging" the tolerances to see if the normal tolerance variations will make the circuit misbehave (or stop working completely).
when i'm not sure what value of a resistor will work best, i run sims and make my changes in a "1-2-5" series, beginning with the most significant digit and fine tuning with progressively less significant digits. for instance i want a resistance that's at least 1k, but no more than 100k, but i'm not really sure where to begin (and probably in too much of a hurry to sit and figure it out on paper, or i'm fine tuning something like distortion levels which are much quicker figured out by the sim). so i start in the approximate middle with 50k and run the sim. then i run the sim with it set at 100k and 20k, and see which one is closer to the desired result. let's say 20k is closer, but not close enough, so i go to 10k. 10k is good, but too far, so i know it's between 10k and 20k, so i go to 15k. closer, but still needs to be higher, but still less than 20k. so i take a "2" step and go to 17k, really good, but can it get better? take an additional "1" step to 18k.... too far, go back to 17k. if i want to narrow it further i go to the next smaller "5" step and go to 17.5k. so i go in big steps of 1-2-5-10, and narrow it down in smaller "1-2-5" steps. some may prefer "1-3-5" or "1-3-7". i actually "borrowed" the "1-2-5" sequence from the range switch sequence of tektronix scopes.
in any case, the sim is really good for "building" the circuit and making sure it will work without "letting out the magic smoke" before i actually put it together. you can see what the device currents will be and compare the currents with the spec sheet of the transistor.