Several thoughts:
As
@helli says the 15 mins is the differential between actual and interstitial. Presumably that is also a generic number, since not only are we all different, but that generic difference must be affected by our degree of insulin resistance at any one moment in time.
However the Libre 2 algorithm attempts to predict where it is, in relation to actual, thus provide a shortened time gap - this is claimed to be as little as 2 1/2 mins, but it feels nearer to 5 or 7 mins for me - with or without Diabox. [Realistically I don't think any tighter estimation of that lag is possible by me from flash scans and LibreLink alone]. Then you have a Diabox algorithm taking the Libre 2 output and making its own predictions! I didn't find any detail about Diabox translation of the interstitial to actual relationship (and I got bored and slightly lost following links when doing that research). So I just accept the principle that there is some lag.
While your BG is changing, then any difference between actual and interstitial is essentially much more complex anyway. Comparisons can only be of value when in steady state, ie horizontal trend or 0.0 mmol/L/min change.
Even when at steady state one set of comparison readings is fairly irrelevant. There is still the "randomness of Diabetes" which can make any one reading or any few hours of readings inexplicable! I have found it takes time and lots of "trial and learning" to get comfortable with interpreting the trend visible from Libre 2 on Diabox.
The Diabox app shows just how much interstitial change is going on minute by minute. And if in the Diabox settings you are updating minute by minute, then after a while the app algorithm struggles to just keep up and the individual readings become less reliable; but the Diabox graph does smooth these apparent blips out into the 5 minute plots.
I still have to remind myself to not "overreact" to an isolated reading - yet still to definitely react to a high downward trend. I have my lower downward alert (in settings / more alerts) set for changes at or greater than 0.3mmol/L/min with both vibration and sound. I just monitor only if the downward trend is at 0.2 or less and then make a judgement if that drop continues every 5 minutes. But I'm retired and can "make the time" to do this; also it's much more difficult to respond or even diligently monitor when I'm driving since I am potentially playing with my phone and thus at risk of being prosecuted ... this is one of those conundrums where it makes sense to head off a hypo while driving but with my tech I'm in the zone of "driving without due care and attention!!"