That rounding rule was created by statisticians. Based only on proximity to the nearest rounded value, a trailing 1,2,3,4 causes the value to go down and a trailing 5,6,7,8,9 causes the value to go up; a ratio of four roundings down to five roundings up, which causes large numbers of roundings to shift the results upward.The reason for the asymmetry is that rounding a trailing zero causes no change in the value.
To fix this problem statisticians decided that since even and odd values are assumed to be random, shifting the effect of the 5 based on even/odd result makes it average out to 4.5 roundings down to 4.5 roundings up, for large numbers of roundings.
When it comes to double dimensioning the secondary dimensions should be entirely within the primary range regardless of the numerical error introduced by the unit conversion and this means ignoring any statistical rounding rules that would allow values to be outside that range.
In this case, I think it was a huge mistake for PTC to allow rounding of dimensions.
As a general FYI
.154 mm = 0.00606299212598425197 inches, approximately and
100mm = 3.937 exactly, not 3.94, so the original conversions are off.
If the goal is to make the range from
100mm to 100.15mm,
the OP desired inch conversion results in a range from
100.076 mm to 100.2284mm (3.94 + .006),
adding significantly (.0784mm) to the tolerance range.
The PTC conversion range result is 100.076 to 100.203 (3.94+.005) , which is still bad, but not as bad.
Had the secondary dimensions been 3 places instead of 2 then .005 tolerance would result in losing .0232mm of dimension range on the top end and using .006 would accept parts .0022mm outside the primary range.
