Effect of a different protection angle (0.3, 0.5, 1.0 λ∕D) over separation from the central star. All values are computed for an observation sequence of 90 min. The color of the line encodes different elevations of the target over the horizon at meridian passage. The line style gives the protection angles. Left panel: total fraction of training data lost due to the protection angle averaged over the whole sequence. Right panel: average temporal separation to the nearest viable frame outside the protection zone. It should be noted that for a temporal systematics model (horizontal lines) that excludes all pixels affected by a companion signal, the exclusion is solely determined by the field-of-view rotation. At 40° total rotation, we would therefore exclude ~10% of the data, regardless of separation, presenting a big advantage at short separations over temporal exclusion criteria.