On this day, the frequency of people who study online for100 ~19 minutes is 0.35, so the number of people is 40× 0.35 =14;
The sample of "40 students" is too small to be representative compared with the overall data of "all students in Grade Three of our school", so it is unreasonable to infer the online learning time of all students in Grade Three of our school only by using these 40 students as samples.
So the answer is: 14 people, which is unreasonable.