To my point of view there is no "good delay" LD50 or LD95 should be specify with the corresponding delay like LD50(48h). Those values may be going up to 96h (maximum observed in publications). Take the most significant.

You have to conduct preliminary expriments for the periods that give the significant results. Please see the following links:

Article Toxicity of Six Ethanol Plant Extracts Against the Green Pea...

It depends on why you are doing this. If the issue is simple pest control then it should not matter much which time interval is used. If the issue is using insect pest management as a tool in disease management then you might be best off using the shortest time interval. If the issue is effects on non-targets, then sublethal effeccts may play an even greater role.

You also need to decide how exact you need your data. Initially, the exact answer is a career killer. The effect of a poison is a function of dose per kilogram. In turn, this is a function of the route of exposure and the acquisition rate for each route. This will be determined by the field application method (soil-foliar, spray-drench, ulv-oil-water-granular-etc...), and formulation effects. You then have a time-dose mortality experiment. That is, the insect encounters the toxicant at some rate (mg/minute), and the problem is to figure out if the acquisition rate exceeds the metabolism rate, the degradation rate of the pesticide, and the level of toxicant that is lethal. It will be made more complex by phenotype differences with varying degrees of sensitivity to the toxicant.

A more typical approach is to use many insects of unknown weight, use mg/l concentrations, and then calculate the dose-response curve for each time interval. If you are ambitious you could take the LD50 and regress that on time. Longer time intervals will have lower LD50. There are very few papers that do a time-dose mortality experiment of any type. As far as I remember all of them run a typical assay (but I have not looked at this literature in a decade). Do a literature search on "time-dose mortality" and see what shows up. You might need the help of a good statistician to write the program that does this analysis. It is not a point-and-click feature of any software package that I am aware of (though I am a decade out of date). The dose-mortality relationship is functionally similar to a time-mortality relationship.

Under laboratory conditions 24 h mortality is the best and standard approach for calculating LD values. However, it will also depend on the type of material being evaluated. If the material is suspected as neurotoxin then 24 h data is the best. However, if mortality is based on growth inhibition or growth regulation then long duration experiments need to be conducted vis-a-vis the life cycle of the insect..

If you recorded the mortality data after 12 h, 24 h, 48 h and 72 h, I think You should calculate LD50 and LD95 for every exposition variant. This is what gives You the answer which one is biologically significant. After all, you have the measurements already.

For aphids, 24 h is enough time for the effect of any bioactive compound or formulation to act. For bacterial toxins however, 48 h or 72 h may be more appropriate because bacterial growth on living tissue may not be as fast as the action of chemical compounds on the same tissue. The bacterial filaments have to grow first on the aphid to produce the required amount of toxin(s) that will eventually kill the insect. The toxicity can also be influenced by the species or strain of the Streptomyces.

There is no definite time since it depends on how fast a test material acts on the aphid. Best time can be determined by conducting your actual tests to find when you reach 100% mortality under the conditions you are testing.

You can then determine the best time point to analyze.

Another way to answer the question is to figure out how fast the material decays in the field. You can then say that I need a residual life of at least 3 weeks in the field to make an economical application. You can then estimate the available dose at 3 weeks, and you could test that dose and figure out the LT50 (rather than the LD50), to see if the product will still be effective at 3.5 weeks. Make sure that you include plant growth as a dilution factor in your calculations, not just hydrolysis/photolysis/(wind and rain fastness) effects.

calculation LD50 and LD95 is usually done at a precise time. that time depend on infectivity potential of the tested strains. A correct choice can be made if you draw graph of agressivity with time. based on this you can choise where strain act differently (i mean where variability of average mortality appear)