The full description of this method is available in this presentation of Uwe Pilz:
It is necessary to say, that this method doesn’t provide raw measures, but theoretical results which are already reduced. This is huge difference against all others visual and CCD observations.
Comet 252P example is for large and diffuse, very gaseous comet. This first examples show the method to failing on such kind of comets, producing magnitudes which are nearly 4 magnitudes underestimated. Apparently the precise CCD measure itself are not impossible for such kinds of comets, classic CCD data of Thomas Lehmann and Ken Harikae fits visual magnitudes very well even for this kind of comet.
Comet C/2011 L4 (PANSTARRS) doesn’t have much observations, the three that was made before perihelion shows a large fluctuations (error reaching almost 1 mag when we are comparing data made by kphot) and a huge overcorrecting of magnitude which is resulting to unrealistic brightness approximately 1 mag over visual magnitudes. I note this comet was very dusty.
Next comet C/2012 K1 (PANSTARRS) shows originally method producing slightly fainter results when after passing perihelion passage the undervaluing of magnitude reach over 1 mag compared to visual data. Later when comet magnitude was decreasing, the kphot results started to approaching visual data and at the end of visibility period, the results become again slightly overvalued. The changes are probably caused by changing of comet dust-to-gas ratio, the comet was more gaseous during perihelion passage, however after the distance from Sun was increased, the dust take major role in cometary coma.
Another good example of dusty comet was famous C/2013 A1 (Siding-Spring). At early stage and before perihelion passage, the kphot results was comparable to visual data, however
Next analysed comet was C/2013 UQ4 (Catalina), originally the kphot results giving good agreement with visual data, however still producing large scatter (a 1 mag difference is unusual for CCD data). Near and shortly after perihelion passage, the kphot results starts to lacking behind visual data for almost 2 magnitudes. There is also a surprising dataset of kphot results in far distance from Sun, which got unusually large scatter, over 2 magnitudes.
Measures for comet C/2013 US10 (Catalina) shows a very good fit in late 2014, while during approaching to Sun, the kphot started to generate magnitudes overvalued by approximately 0.5 mag compared to visual magnitudes. Around perihelion passage, the kphot results starts to running behind visuals again and resulted to values which are undervalued by 1 mag compared to visual magnitudes as the gas component starts to play major role in comet brightness.
The same effect seems to be visible on several datapoints for comet C/2013 X1 (PANSTARRS), overvaluing the magnitude when comet is far away from Sun and undervaluing around perihelion.
Another example of gas rich comet is C/2014 E2 (Jacques), in this case kphot generate slightly undervalues results compared to visual data.
Data of comet C/2014 Q1 (PANSTARRS) shows a great consistency of kphot results with other CCD magnitudes. But there are few erroneous points again, which are overvaluing the magnitude by 1 mag and more.
Next example of very gas rich comet is C/2014 Q2 (Lovejoy), the results from kphot giving usual view, originally good fit which later change to extremely large undervalue of magnitude against visual observations or classic total coma CCD measures.
For comet C/2014 S2 (PANSTARRS) the kphot is giving results near the top visual estimates, however near perihelion, its results was even overvalued by almost 1 mag compared to visual magnitudes. It is a question if this comet was mostly dominated by dust or gas, as there was visible both, large outer coma and also a bright dust tail.
Last comet C/2015 G2 (MASTER) is again an example of typical gas rich comet, and also the kphot exhibit typical error for this kinds of comet, as in large distance it produces nearly same results as visual but closer to perihelion the error starts to grow to nearly 2 magnitude smaller than visual magnitudes.
Many comparisons of kphot results calculated by Kevin Hills by his data shows, that the coma model used by kphot to correct magnitudes is too simple. Apparently most comets have too unique coma profile for one simple fit. The method fails for gas rich comets, in which cases the correction is not enough, while for dust rich comets, it can produce values that are overcorrected and higher than the real comet magnitude. Which is most sad is that the error of this method isn’t consistent in various heliocentric distances for individual comets as the gas and dust contribution to coma brightness is changing due different heat income to nucleus. Also there seems to be small error, when comets are again distancing from Sun, caused by contribution of slow moving dust grains.
The few very erroneous points, which exhibits a large (1-2 mag) overcorrecting of magnitudes seems to be most likely product of bright stars inside coma. The original error caused by presence of such stars in multibox results is later multiplied by kphot to ridiculous values.
Both aspects very complicating the use of this method, as we can easy see on the examples, the data reduced by kphot is producing very inconsistent data sets with unreal trends compared to visual or other CCD datas. The problem is also that the data cannot be simply reversed to original measures. This excludes use of this kind of data in any serious light curve analysis, there is danger that this data can cause very unrealistic brightness in large distances and also can manipulate the brightness trends in various heliocentric distances to unreal results. The various range of possible errors also excluding any other corrections of kphot results.
It is a big question if it is even possible to find a solution on the effects noted above, apparently every comet got different, and dynamically changing coma profile in wide range of heliocentric distances.
For observers who are doing multibox measures I suggest it is better way to provide raw data for different aperture sizes (so there would be 6 instead of 1 measure, each for multiple different boxes) so the correction can be done later and for various coma models.
Still the best way to provide most accurate measures are to do the classic CCD measure of correctly exposed total coma diameters, which is of course more time consuming, but the value of such data is extremely high (the data are mostly comparable with visual magnitudes, but provides very small scatter within 0.2 - 0.3 mag). I can suggest a good presentation of Thomas Lehmann who is very successful in providing correct measures: https://drive.google.com/open?id=0B8OVnW0WLB3EWHdBVVN5UHJyX3M