Take a closer look at the actual original strings. The second two strings match multiple substrings, so the desired output actually shows this, but in a sort-of-confusing way. Take this example:

STRING
-------
GCGCTCGACGC  

SUBSTRINGS
----------
GCGC ACG    ==  [GCGC]TC[ACG]C
[download]

But, when we have OVERLAPPING sequences the output should 'mash-up' a bit:

STRING
-------
GCGCTCGACGC  

SUBSTRINGS
----------
GCGC GCTC  == [GCGCTC]GACGC
[download]

Do you see how GCGC AND GCTC MERGE into one single substring for the desired output?

So I think the algorithm should look like this:

Make as many straight matches as you can

If your match is within a string that has already been matched, modify that match to include the new match

Can you imagine how messy this would look if you had 100 substrings and a main string running 10,000 letters long (which I assume is possible because this stuff looks like gene sequence data)?

Celebrate Intellectual Diversity

Comment on Re^2: Bracketing Substring(s) in the String Select or Download Code