Not sure if I understood this statement correctly, so let me go over ISF. Apologies if this has all been clear already: a high ISF [(mg/dL)/U] means that it takes less insulin to correct for a bg that is away from a target. As an example, suppose that Loop predicts eventual bg = 150 mg/dL, that the target is 100 mg/dL, and that ISF = 50 (mg/dL)/U. Then Loop would calculate that (150-100)/50 = 1 U is required as a correction, and would issue a temp basal of +2 U/h above the normal basal rate, in order to deliver the required correction over the next 30 minutes. If, however, ISF were higher, say ISF = 100 (mg/dL)/U, only 0.5 U would be required, and the temp basal command would be +1 U/h. In general, the higher the ISF is, the less aggressive Loop (or OpenAPS) is. Conversely, a lower ISF results in more aggressive operation of the closed-loop system.
A good starting point for ISF would be the value you've used for manual corrections. With Loop, one approach I've used to make adjustments was to look at overnight performance first. A too low ISF tends to result in overnight oscillations - high temp followed by low temp followed by high temp etc. Fortunately, the resulting bg oscillations tend to be relatively small in amplitude, so that's not so bad, but it's a good way to see that ISF is too low. A too high ISF tends to result in bg staying away from the target for too long. A good ISF should result in bg soft landing into target range overnight. In my case, I moved ISF up compared to what I thought was a good value prior to looping. The above description assumes the overnight basal rate is generally set correctly, which can be verified by taking notes of how much extra insulin the Loop delivered (or subtracted) overnight (over at least a few days - I'd not rush to make adjustments to default basal rates).
DIA is also important for Loop operation. I started with a shorter (3 hour) DIA, which I had used in manual operation. Extended that to 4.5 hours, and have had better overall results since. As usual YDMV, but I do not think insulin absorption is that much different from person to person. A DIA of 4 hours or more seems to be more consistent with published curves for the fast acting insulins available today.
Another important factor and (in my experience) the most difficult one is the carb absoption time, which of course depends on the meal type, but also varies a lot from person to person (and time of day, etc), so it is virtually impossible to give any general guidelines. For example, the Loop defaults of 120, 180 and 240 minutes are not working so well for me. For fast carbs, I'd use 90 minutes, while for a complex meal I'd enter a carby portion as 120 min and the rest as up to 360 min. I'd bolus for the first part, and let Loop high temp for the second (which is where max temp basal setting also plays a role). I should note that OpenAPS algorithm called oref0 deals with carb absorption differently.
When I started looping (OpenAPS, then Loop) more than a year ago, I first focused on the nighttime, and I made adjustments to the overnight basal rates and the ISF to get consistent performance. During that time, I kept the system mostly open-loop during day, and continued to surf manually, with closed-loop tests done whenever convenient - trying out different meal absorption times and different carb ratios. It took me a few weeks to get the settings and the absorption times to work well for me.
I'd not be discouraged by some setbacks - there is a lot to learn and do with DIY closed-loop systems. The experience is very different from what we've been used to do manually. Best luck!