Yeah, but theories based on approximate conclusions can sometimes (and have been) biased in that you are specifically trying to look for something in an experiment to match a theory and often ignoring somethings that are staring them right in the face.
Theories made in this way are bad;
Sit at table --> think of theory --> play with maths until numbers work --> design experiment to validate theory (if possible) --> look for data that agrees with theory --> "prove" theory.
then 5-10 years down the line someone will come along and most likely dis-prove the theory with data from another experiment or by re-analyzing the same data in a different way.
This is the proper way to make theories;
Design experiment to investigate a known phenomena --> collect data on EVERYTHING --> look at data --> create theory --> test theory on new experiment
-- the two ways work best together in practice, no major experiment will be built without some possibility of finding/proving something new.
I'm an experimentalist at heart, I like to do the experiment and analysis
"blind" in that I look at anything and everything that may be of interest and then try to understand it. This is the major problem in building an experiment to just test a theory, there is so much pressure and so much expectation to find what you want, the actual science ends up being poor. Sometimes it just as important to dis-prove a good theory as it is to prove one.
I'm not saying all theories based on "pure" maths are bad, some have been proven to be really sound, but most of the "important" theories have also been based off data or known principles when they are been derived.
This theory of previous universes is nice and all, but we'll never be able to prove it.