Had some success with reconciliation today. Assuming you download the correct latest release, the seed reconciles when using the mnemonic length as the default. I was changing it to 24 words and using that criteria, the output differs. Becoming more confident in the tool using the default length for dice roll 1-6. More testing tomorrow just in case though. Sing out for more enquiries

Not a github user but someone may want to report this against the coleman tool even if someone can just explain it. When using the offline BIP39 tool (iancoleman/bip39, tested on release 0.5.6) with a custom entropy value entered manually, selecting a fixed "Mnemonic Length" from the dropdown (e.g. "24 Words") produces an incorrect result. Specifically, the Entropy, Filtered Entropy, Raw Binary, and Event Count fields all remain correct and consistent with the entered value, but the Binary Checksum field is computed incorrectly (verified by independently calculating the first 8 bits of SHA-256 of the entropy by hand and with a separate reference BIP39 library — both gave a different, correct checksum). This incorrect checksum then cascades into completely different Word Indexes and a completely different final mnemonic phrase, unrelated to the entered entropy. By contrast, leaving the "Mnemonic Length" dropdown on its default setting, "Use Raw Entropy (3 words per 32 bits)," computes the checksum correctly and produces a mnemonic that matches independent verification exactly. This was confirmed as a genuine bug rather than user error or a stale-page issue by reproducing it twice with two different entropy values, and by verifying the downloaded HTML file's SHA-256 hash against the developer's official published release hash, confirming the file itself was unmodified. The workaround is to always leave the Mnemonic Length dropdown on "Use Raw Entropy" when supplying custom entropy manually, rather than selecting a fixed word count.