You have two proteins with weak, remote similarity. You know their sequences. You also know the sequence of the original DNA. Why would you expect a better alignment using the protein sequences
Answers
Since there are just 20 amino acids associated with synthesis of protein, however there are possible 64 codons, there is a variable level of degeneracy.
For instance, despite the fact that Met and Trp are encoded by exceptional codons, Arg, Ser and Leu each have six. Given diverse content and codon preferences of GC in various life forms, you may have a much lower dimension of sequence identity at the nucleotide level than at the protein level. Along these lines, when searching for related successions in the database, a protein search should be more delicate than a nucleotide data search. Clearly, this applies for coding region: for non-coding districts, the determination pressure / selection pressure keeping up sequence homologies acts at the dimension of the nucleotide sequence.