I actually use paint shop.  I'm just never 
bothered to invest the time to learn photoshop.

theredhood had the basic idea.  You just scan 
different parts of the cel and then copy & paste 
them together.  It's very difficult to do this if 
the cel or background does not have one flat edge 
to align the different parts on the scanner bed.

For most cels, I usually scan at 400-600 dpi and 
then shrink the image.  For cels that require 
more than one swipe on the scanner, I won't 
shrink the image until I copy and pasted all the 
parts together.  These scans do get very large, 
so a machine with lots of ram is very handy.  I 
prefer to shrink large scans (instead of just 
scanning directly to a certain size) since it 
gives a more accurate representation of the cel 
and hardly requires any touch-ups.  Also, it 
helps to correct inconsistancies (like small 
light changes that cause the appearance of 
contrast bars).  In addition, less than perfect 
alignments disappear more easily when shrinking a 
large image.

Here's what I did for the Mayuka cel.  My scanner 
bed has an area of 8.5 x 11.5".  The Mayuka cel 
is 10.5" x 20".  By placing the cel so that the 
short edges are parallel to the scanner bed's 
long edges, it will require 3 swipes: the top, 
middle, and bottom part of the cel.  Luckily, the 
cel's long edges are flat, so it will be easy to 
align scans.  In this case, I just made sure that 
only the cel's left long edge was flush against 
the scanner bed's left short edge, and ignored 
the alignment of the other sides of the cel.

Next, I just positioned the cel over the scanner 
bed and took a scan for each of the 3 parts.  I 
then cropped each part so that there is little 
overlap between the parts.  (It's important to 
crop the edges for non-BG cels because of 
shadows).  When copy & pasting large scans 
together, make sure your machine has lots of ram; 
otherwise, it may freeze during a copy & paste 
(and lose your scans; so save the scan parts 
before doing this; if necessary, scan at a lower 
dpi).  Before copy & pasting one scan part into 
another, make sure you enlarge the image area 
beforehand.  Alignment of the different scan 
parts will take a while for a large image, so try 
not to hurry.  It's ok if the alignment isn't 
perfect since shrinking a very large image will 
help make it less or even unnoticeable.  For a 3 
part scan, you'll have to do this twice.  
Afterwards, I cropped any extraneous garbage 
around the scan and shrinked it.  My combined 
scan was over 9000 pixels tall before I shrinked 
it to 1525 pixels (yeah, I have lots of ram).  
That degree of shrinking is probably why you 
can't see where I fused the different parts 
together (and why I don't have to do any touchups 
like sharpening).

I'd say my scans are pretty decent for a ~$80 
scanner, heh.  Later.