Understand Odd numbers in custom designed RSM+setting continuuos factors to zero

[#1]

Dear JMP-Experts,

I have two questions:

Question 1) I created two RSM-Designs using the custom Designer
-The first design has 3 Factors, of which two factors (Factor1 and 2) are continuous with three factor settings  and the third factor is discrete numerical with four factor settings (two include 0 as additional setting)-->See below.
-The second design, is similar to plan 1 but with four factors; three factors continuous and 1 factor discrete numerical to include 0 (See below)
-->Now my question: In both cases, JMP creates some odd numbers as factor setting (See below in bold), which are hard for me to explain? What is the mathematical and pratical reason for these odd numbers and how are they generated and how is the factor chosen with these odd factor setting (seems to be a bit random)

Plan 1

#Run	Factor1	Factor2	Factor3
1	252	12	0.01
2	280	10	0.01
3	280	10	0.02
4	308	12	0.02
5	280	8	0.02
6	290.92	10	0.03
7	252	10	0.02
8	264.04	12	0.03
9	252	8	0
10	280	8	0.01
11	308	8	0.03
12	252	8	0.03
13	308	10	0.01
14	308	8	0
15	289.52	12	0
16	280	10	0

Plan 2

#Run	Factor1	Factor2	Factor3	Factor4
1	308	12	0.03	50
2	280	10	0	50
3	252	12	0.03	55
4	308	12	0.01	55
5	252	8	0	45
6	280	10	0.02	55
7	252	8	0.01	55
8	252	12	0.01	45
9	280	8	0.03	45
10	308	10	0.02	45
11	308	8	0.03	55
12	252	10	0.02	50
13	252	12	0	55
14	252	8	0.02	50
15	308	8	0.01	46.85
16	280	10	0.01	50
17	280	10	0.01	50
18	280	12	0.02	50
19	280	12	0.02	45
20	308	12	0	45
21	308	8	0	55
22	252	10	0.03	45

• Question 2:
  Why is it not a good idea not use 0 as factor setting for a RSM with continuous factors. I know that a factor setted to zero is techical not continuous anymore but rather categorical. But what is the risk of doing that? How can I explain to a non DoE expert not to do that?

Thanks a lot for your help in advance!

[#2]

The Custom Design platform is based on the coordinate exchange algorithm. It searches for factor settings that maximize a design criterion. The RSM button in the Model outline changes the criterion to I-optimality, which minimizes the integrated variance of prediction

Some solutions are not intuitive or practical. You can change the level to the midpoint. This change will lower the optimality but to a very small extent.