I find that I can make a surprising variety of machinery work. The fact that fit follows a certain fashion indicates that there is more than one perfect frame. In the 1940s, they'd have put me on a 63 or so frame, in the 1980s, it was 57 (smallest frame possible with "Campy seatpost to the limit"), and today, I am afraid they'd put me on a 53 cm with sloping top tube! And honestly, I would be comfortable on all (except maybe the modern one, which I haven't tried).
My old bikes aren't all perfectly sized...
On a bigger frame, I just pretend you live in the 1940s, and lower the seat as far as it goes. If the frame is bigger than that, nothing you can do... If it's smaller, I find it's OK to a limit. I just tested a bike for VBQ that, in a 57 cm frame size, actually measures 54 cm c-c. It rides fine - seat and head tubes are extended upwards, and it feels like my 58 cm Rivendell (which has a sloping top tube), which feels similar to a 61 cm Singer (classic frame geometry) fit-wise. Considering that I used to race a classic 57 c-t frame with no problems and today prefer a 61 c-t shows that there is some leeway. Once it gets much too small, it's not so great - with the exception of the above-mentioned test bike, I have had unsatisfactory experiences with frames smaller than 57 c-t.
Another exception was our PBP tandem: It was smaller than would be ideal, and the bars were too low (direct clamp stem), but randonneur bars helped a bit. No real problems, either, considering we rode the thing for almost 53 hours.
Stems that are too short aren't a problem, either, as I just sit a bit more upright on those bikes. It doesn't even seem to slow me down - some of my best times over long distances were on bikes where I would have spec'd a stem that was 2 cm longer... On bikes where the reach is too long, I raise the bars, which brings them closer. But that is less ideal - fine for short rides, but for really fast, long rides, I take other bikes.
As Richard pointed out, when getting a new bike, getting the best fit possible is a good idea. But to think that half a cm here or there will make a huge difference just isn't true for most people.
I even wonder about saddle height within reason. Mine has gone down a good 1.5 cm in recent years! In the 1949 technical trials, two riders took turns on the same bike. To save time, most didn't bother to adjust the saddle height. Obviously, they chose riders with similar inseams... In my experience, a tad too low is better than a tad too high. Others' experience may vary in this respect.
Mike's experience indicates that Q factor shouldn't be overlooked. Can't comment on that - for me, Campy NR and all the other old ones, which are similar, work best. Racers used to believe low Q was paramount.
Finally, for me crank length is important. I can handle anything up to 172.5 mm, but beyond, I just don't seem to have the same spin - despite leg length formulas indicating I should be on at least 175 mm. Of course, with most older bikes equipped with 170s, it's not a problem for me.
In the end, I am astonished by the number of bikes I can ride that
don't look much alike size-wise. It is true that I have a few perfect
bikes that I prefer. But riding 400 km on a bike that is a tad small
in every direction isn't a problem at all - see the old story and
photo at
http://www.mindspring.com/