perl, 205 191 189 characters/ 205 characters (fully implemented)
Some parts were inspired by the earlier perl/ruby submissions, a couple similar ideas were arrived at independently, the others are original. Shorter version also incorporates some things I saw/learned from other submissions.
Original:
$k{$_}++for grep{$_!~/^(the|and|of|to|a|i|it|in|or|is)$/}map{lc=~/[a-z]+/g}<>;@t=sort{$k{$b}<=>$k{$a}}keys%k;$l=76-length$t[0];printf" %s
",'_'x$l;printf"|%s| $_
",'_'x int$k{$_}/$k{$t[0]}*$l for@t[0..21];
Latest version down to 191 characters:
/^(the|and|of|to|.|i[tns]|or)$/||$k{$_}++for map{lc=~/[a-z]+/g}<>;@e=sort{$k{$b}<=>$k{$a}}keys%k;$n=" %s
";$r=(76-y///c)/$k{$_=$e[0]};map{printf$n,'_'x($k{$_}*$r),$_;$n="|%s| %s
"}@e[0,0..21]
Latest version down to 189 characters:
/^(the|and|of|to|.|i[tns]|or)$/||$k{$_}++for map{lc=~/[a-z]+/g}<>;@_=sort{$k{$b}<=>$k{$a}}keys%k;$n=" %s
";$r=(76-m//)/$k{$_=$_[0]};map{printf$n,'_'x($k{$_}*$r),$_;$n="|%s| %s
"}@_[0,0..21]
This version (205 char) accounts for the lines with words longer than what would be found later.
/^(the|and|of|to|.|i[tns]|or)$/||$k{$_}++for map{lc=~/[a-z]+/g}<>;($r)=sort{$a<=>$b}map{(76-y///c)/$k{$_}}@e=sort{$k{$b}<=>$k{$a}}keys%k;$n=" %s
";map{printf$n,'_'x($k{$_}*$r),$_;$n="|%s| %s
";}@e[0,0..21]
