Three possible solutions...
Put a sand bag on it.
Tell him the program output is theoretical and it really doesn't lift off the support.
You don't want to know the third.

Come on man. 0.1mm? This stuff is not fabricated under a microscope.
It is not justifiable, as
the piping fabrication and fit up is not done to anywhere near such a precise tolerance.
Neither is the structure holding up the spring can.
You cannot adjust the spring hanger to such a precise setting.
you will have much more soil settlement under the support's foundation.

Learn from the mistakes of others. You don't have time to make them all yourself.

Ah but BigInch the computer says....... so it must be right!!!!! Typical case of where the "user" must use his experience. I have seen similar instances where the Stress Analyst (loosely named) adds a 0.1mm gap at a guide to get equipment loads within acceptable limits as a larger gap is not acceptable and then hopes the Construction installs this 0.1mm gap. Again assuming that we are dealing with Rolex watches!!!!

Well then it would seem that the real problem is that people should not be allowed to push buttons before they've tightened a few loose screws.

you must get smarter than the software you're using.

BigInch,
Had a chuckle but yes that's spot on!!!!

All kidding aside, it is truly imperative that one must know the practical limitations of theoretically based algorithms and software and when and how to apply solutions derived from such to real world problems. In some cases even popularly accepted algorithms, such as the ASCE and American Lifelines soil friction models, cannot be used across the board in all soils. Unconservative errors of up to 200% can result in many circumstances that a novice would think a perfectly applicable algorithm.

you must get smarter than the software you're using.

Software publishers should in fact say more about these things, rather than to sit back, sell as many copies as they can and rely solely on their blanket disclaimers. It's getting to the point where that behavior is no longer acceptable.

you must get smarter than the software you're using.

BigInch,
Totally agree with you the Technical Reference material of some software is diabolical but as you say the Software publishers "hide" behind the disclaimers - even if there are "bugs" in the software they pull the "Users should ensure the results are valid" statement.

Yes and I've also seen plenty of user manuals that don't tell you what algorithms the program is using, so the user cannot be sure the algorithm is valid at all, as many programs don't check input data for validity. The only way to verify is by comparison to experience, or another program's output. Another bunch of programs will tell you the algorithms in general, but not enough detail in how, or what order certain equations are applied for the user to duplicate the results. That is, if the calculations can be done by hand at all. There is no substitute for experience ... at times.

you must get smarter than the software you're using.

edwin, why not use less expensive non-spring support(s) instead of a spring hanger given that deflection is so small? Significant load redistribution can occur before a pipe fully lifts off a support, so the calculated .1 mm deflection is not necessarily a concern.. Your allowable loading seems low for 20" pipe. To my knowledge, API 537 doesn't provide allowable load values. Are you using API 560?

Regarding checking program "algorithms", other than a feel for how the system should be displacing, how would you go about rigorously verifying those algorithms for real world piping layouts beyond a simple L configuration? Especially with NL pipe supports. And even if you had each step of each algorithm, you'd need to account for a slew of other related assumptions and limitations of the program.

To fix it at the CAESAR II (assumed) level, change the "+Y" (assumed) support nearest the site of the 0.1 mm upwards displacement to "Y". Alternatively, artificially impose a "-Y" restraint with zero gap at the nozzle or support site, and state in the cover page of the calculation file output why you have done one or more of the above.

One fact that I have discovered is that some clients want to see that computer-generated (paraphrasing) "Code Compliance Check Passed" or "Nozzle Loads Acceptable" message in the output; otherwise, no amount of superseding statements or applications of judgement on the part of the engineer can convince them that things are OK. What has worked for me is to force the program into providing that desirable output message by stating some qualifying assumptions up front, where they are innocuous and can be wordsmithed to give the appearance of intelligent thought. In other words, as long as you kinow what you are doing, cheat.

Are we 100% confident on software to say that 0.1000000 mm liftoff will actually occur? What if its zero mm; no problem then. What if its one whole mm? Load redistribution and local yielding are keywords then, caus they will, eventually after a few cyclus, make sure no liftoff will happen again.
Using one eight of an inch as a max allowable liftoff is not uncomcmon. Keep in mind the tolerances we have on fabricated pipe; as BI rightfully pointes out, we dont make them using microscopes.