I don't know if this issue is still extant. I'm the developer of conversion tool that was discussed. diCrunch does produce fully accurate conversions between most diacritic systems and Devanagari script. However, most of the transliterations you read online don't have the necessary diacritic marks to accurately reproduce the original in an Indic script.
The Devanagari versions you find in Wikipedia are accurate. The "anitya" of Sanskrit mutates to "aniccaa" in Pali, and the vertical bar you see at the end stands for a long A, and same applies for anatta. Written in Harvard-Kyoto transliteration, they are dukkha, aniccA, anattA
, the proper IAST transliteration is dukkha, aniccā, anattā
, and the Devanagari is दुक्ख, अनिच्चा, अनत्ता.
What happened in your case is you wrote Dhamma, and had "Harvard-Kyoto" selected. In Harvard-Kyoto, the diacritics are marked by capitalization, and as such "D" is not the same as "d", "A" does not equal "a", and so forth. This is why you need to have accurate transliterations (or otherwise know your grammar!) if you want to render Romanized words into Indic script. At least before you tattoo them on your arm.
If you wanted to be true to the roots, you'd probably write it in . This is the script you'd have seen in the edicts of Ashoka, if you've read the small print on your pilgrimages. is of course another very classic (and cool-looking!) script, the typeface of Tipitaka as the Theravadins of yore had it in Sri Lanka.
I'd actually love to add Brahmi and Sinhalese to diCrunch when I get a bit of spare hobby time. For one, I'd love to print out some suttas in the "original" and see if I can decipher them! That'd surely yield a mighty dose of nirvana.