The bilinear method often can't fit one of Jmax and TPU. But we could still estimate a lower bound for these values. e.g. for TPU, if it is not fitted, we nonetheless know it must be at least
TPU > (A + Rd) / 3
for the largest value of A in the curve. If estimated TPU or Jmax is NA, can we return eg TPU_lb which is lower bound for TPU?