Three attested filters, evaluated against your recorded birth time — each reporting its own verdict beside its own citation.
This page does not return a rectified birth time. It takes the time you already have and tests it against three attested filters, each reporting its own verdict beside its own citation, its own confidence grade, and the finest interval over which that verdict can change. There is no score, no ranking and no best candidate, because no text prescribes one.
Read that as the design rather than as a shortfall. Three passes do not confirm a birth time and three failures do not refute one. The rows render in the engine's fixed order, they are not sorted by anything, and nothing on the page is counted, combined or weighted. Each row is meant to be read on its own terms, against its own source and its own resolution.
The Kunda correction comes from Prasna Marga chapter V, topic 52, stanzas 8 and 9. Multiply the ascendant's sidereal longitude by 81 - the text's own glossary defines Kunda as the number 81 - expunge whole multiples of twelve signs, and read the nakshatra the result falls in.
That kundagata star must be your birth star or one of its two Anujanma stars, which the glossary defines as the tenth and the nineteenth from the birth star. If it is not, the ascendant is shifted by one asterism's worth for each asterism of shortfall.
One asterism of kunda is 800 arc-minutes divided by 81, which is 9.877 arc-minutes of ascendant; Raman prints the rounded ten, and the page reports both figures. The shortfall can never exceed four asterisms, because the three accepted stars sit exactly nine apart in a 27-cycle, so the largest correction Kunda can ever apply is 39.51 arc-minutes of lagna - between roughly 1.8 and 3.1 minutes of clock time at a real Delhi ascendant rate.
That derives the tradition's three-minute limit from geometry, and it is also the warning: given a larger error Kunda does not fail loudly, it locks onto the wrong bracket and returns a confident answer.
The second filter is B.V. Raman's Rule 1 from the birth-verification chapter of Hindu Predictive Astrology: take the time of birth in vighatis, multiply by four, divide by nine, and the remainder must give the ruling constellation counted from Ashwini, Magha or Mula. One vighati is 24 seconds. Because four and nine share no common factor, the remainder changes with every single vighati and cycles through all nine residues, so the accepting window is exactly one vighati, 24 seconds, recurring every 3.6 minutes.
Two print editions of the same book disagree about the unit, one saying ghaties and the other vighatis. This app adopts vighatis and states why in checkable terms rather than editorial ones: the rule needs an integer remainder in order to count from Ashwini and ghatis give a fraction, the edition that says vighatis is the one carrying a self-consistent worked example, and ghatis would give a 54-minute granularity that is useless for the purpose the chapter states.
The disagreement is reported rather than hidden. The filter is carried at medium confidence, being one modern author with two conflicting editions and no second transmission, and it is not BPHS.
The third filter is Raman's Rule 3: the ascendant should be the 5th, the 7th or the 9th sign from the sign occupied by the lord of the Moon's sign, or in some cases the Moon's own sign. The page marks it indicator only and never lets it act as a filter or a tie-break, and that restriction is enforced in the engine rather than left to the reader.
Three independent reasons force it. Raman gives no worked example for this rule in either edition, unlike his Rules 1 and 2. The accept set grows by escape clause, naming the 5th and 9th, then adding the 7th, then adding in some cases the Moon's own sign.
And with that clause counted in it accepts four signs out of twelve, a third of all charts by chance, at sign resolution, which is a roughly two-hour ascendant window. It can flag a lagna grossly inconsistent with the Moon's dispositor; it can never settle a birth time.
The obvious next move is to count how many rows passed, or to weight them, and name a best candidate. This app refuses to. The scoring weights appear in no text, and the rows are not statistically independent - they are all driven by the same candidate instant - so a number here would be inventing Jyotish rather than computing it.
Rectification from dated life events is refused for the same reason and is offered nowhere in the app. Every individual criterion is citable, but with eight to eleven booleans per event swept over hundreds of candidate times, the top scorer is usually an artefact of multiple comparisons rather than a signal. Without a ranking there is no answer to return, and the ranking is exactly the part no source licenses.
The refusals are rendered on the page as data, each with its reason and its source, so that a gap reads as a decision rather than an oversight. Tattva-shodhana is refused because its two apparently corroborating sources are one modern book reaching us twice through one translator, because no acceptance criterion links a tattva to any observable chart feature, and because a 90-minute cycle with no observable anchor claiming 12-second accuracy would inject false precision into every dasha and varga downstream.
Raman's Rule 2, which compares the Bhava Lagna's sign to the ascendant's, is refused on measurement rather than on principle: the two agree only 63.1 per cent of the time at 28.6 degrees north, so the test would falsely reject 36.9 per cent of correct birth times, and above 40 degrees north it rejects the majority.
Its own maximum deviation is a full sign, the same magnitude as the one-hour error it claims to catch. The 225-vighati gender table is refused because no pre-modern source for it could be located.