No, he's right - you sequence the amino acids and count the matches. I've done runs on proteins from organisms as different as E. coli and rhesus monkeys. This does not, however, measure secondary and tertiary structure, which is what you really need if you're to determing biological activities, but the amino count won't get you there by itself.
No. Run some blasts or simple seq aligns with disparate proteins. You'll not get the right matches in many many cases.
Some random sequence might have a high match in an unconserved region and skew the alignment. And the differences in size always make it hard to know where to begin the alignments, where the gaps are.
heck, run programs with slightly different algorithms from different labs and you'll get different alignments and gaps etc...
Of course it is easy to line up two sequences and count matches. But if they differ by a lot there is a lot more to it than just slapping them next to each other.
Take for example a hypothetical protein with two transmembrane domains. One is in bacteria, the other mammals. The latter is 400 aa and the former 250. Line them up from their amino terminals and look at the matches -- it means nothing. I shouldn't have to have to tell you this.