The market report and the rent roll disagree completely, so which is lying

Your arithmetic is right and so is the concern. Six weeks free on twelve months is 11.5% off gross, two months on a fifteen-month term is 13.3%. If a submarket went from zero concessions to those levels over a year, net effective rents fell far more than the asking series shows. Plenty of published series are asking rent, some are effective, and the two get quoted interchangeably in market commentary. Check the methodology note on whatever report you're using before you compare it to anything.

The sequence in a real recovery is fairly consistent. Concessions compress first, because they're the fastest lever a leasing office has and the first one pulled back when traffic improves. Asking rents move later. So a submarket where asking is flat and concessions are still widening has not turned, and a submarket where asking is flat but concessions went from eight weeks to four is well into recovery even though the headline looks identical.

What I'd track at property level: new lease trade-out and renewal trade-out separately, both net of concessions, plus concession weeks offered as its own series. Renewal spreads stay positive long after new lease spreads go negative, because in-place residents are anchored to what they already pay. A blended number hides that. Add occupancy plus the exposure figure (occupied plus notice-to-vacate against total units), since a property can hold 94% occupancy while exposure tells you 11% of units are leaving in sixty days.

The fifteen-month lease in your comp set is worth thinking about beyond the discount math. Extended terms are a way to push expirations out of the next soft leasing season, which flatters near-term performance and stacks a heavier roll into a specific window later. When you underwrite that property you're inheriting the expiration calendar someone else built, and if a large block rolls in one quarter you renegotiate at whatever the market gives you that quarter.