The problem with the Gibson calculator is that it seriously overestimates the strength of passwords.
The issue in the OP is whether concatenated passwords are secure. The fact is that a determined adversary will try anything that gives them an advantage over a dumb brute force search.
That means using dictionaries, and algorithms that generated "common" patterns based on dictionaries. There have been many "hacks" in the past resulting in large numbers of passwords being compromised - there are whitehat hackers, password recovery companies, etc. who have obtained these databases and used statistical analysis to work out just what dictionary words are most likely, what patterns of "obfuscation" are used (e.g. capitals, l337spk, character substitution, reversed words, concatenation, etc.) most frequently, etc. You can bet, that if "whitehat" researchers and commercial enterprises are doing this, that the "blackhats" are also doing the same.
Now, common hacking tools might not do this, but remember that "pro" hackers write their own tools and do their own research.
I can well imagine that if a determined adversary has compromised some of your passwords, the first thing that they will try is various corruptions or combinations of those.