Interesting post, thanks for including the emergence as well as the finished product. My first thought was rather different: I'd partition the data, and recombine the results of simple separate sorts:

# simple and effective:
@sorted = (sort(grep!/:/,@data), sort(grep/:/,@data));

# or to avoid grepping @data twice: (probably not much gain)
my (@colo, @noco);
/:/ ? push @colo, $_ : push @noco, $_  for @data;
@sorted = (sort(@noco), sort(@colo));
[download]