"Sorry, I still don't understand your process."

No, I'm sorry; I'm not explaining it well. The current price doesn't factor into the estimate of IV. The IV estimate is simply the long term trendline of P versus BV, at any point along the 60-year data series of P versus BV. Of course the trendline is actually price vs BV, not IV vs BV, but the assumption is that over the long term IV moves with BV, just as price does, and that the P versus BV trendline provides a reasonable estimate of the IV versus BV trendline.

John Kish estimated IV from Dec 1981 through Dec 2014. He calculated several IV's each year, which he labeled optimistic, conservative, etc., corresponding to the assumed discount rate applied. Each of the calculated IV sets, optimistic, conservative, etc. when plotted against BV tracked BV perfectly (r^2 = 1.0), but they were offset from one another. So which set was closest to actual IV. I postulated that if one used price as an imperfect proxy for IV, and plotted P vs BV, that that trendline might be closer to actual IV than the trendlines using Kish's various sets of assumptions. The P vs BV plot worked, with an r^2 of 0.98, so I used the P vs BV trendline as an estimate of IV.

As you point out, other valuation models also work well. A simple two-column model, as used by Tilson, gave an IV as of Dec '24 of $718K/A-share, as compared to the P vs BV model IV of $724K. I think that one of the best models is Buffett's highest P/BV paid in stock repurchases, and those P/BVs (highest repurchase price = 1.57x BV) are in agreement with the P vs BV trendline model (estimated IV = 1.59x BV). I have no doubt that your two and a half column model is one of the best models. My use of the P vs BV model (which I call the weighing machine model) was not to imply that that was the best model; it was only used because it gave a consistent set of IV estimates usable in a discussion of the trend in IV/BV.