*IF* I understand you correctly, we are the creators (or co-creators?) of the conceptual constructs (or thought forms) that you say are the "carriers of experience", and these carriers of experience are probabilistic in terms of their particle nature, and simultaneously have independent, objective existence in terms of their marble nature?
Here is a bigger picture to explain it in wider context:
1) PMR was evolved. If you would take the Big Bang as starting point, then at the beginning there was a rule set and initial parameters (energy or information seed if you like). This initial information seed / parameters interacted and evolved under given rule set constraints as probabilities. You know, rule set allows certain variations in interaction outcomes, so instead picking one from possible outcomes at each delta t and rolling forward from chosen state, probabilities were combined to create more complex probability functions. That's roughly a mechanism how you can evolve your way up from probable particles to probable stars and probable solar systems etc. (models of those with some uncertainty level in exact values of their parameters at all levels). So far all real and stable in concept but "fuzzy" in exact details.
It's possible but not necessary that this evolution experiment was tampered with, i.e. when too much uncertainty built in the systems, there was some narrowing down steps (choosing the best possible state and continue from it). I don't know.
You may picture the PMR VR as field of mutually influencing probability functions formed from information (consciousness) fractal. Or if you're familiar with software, picture it as formed from Lisp code (in this programming language data and code are the same thing) so you compute on code (mathematic formulas) that will produce result when fed with data (past records from actualized reality database) send then to rendering streams. This code forms a fractal structure, so you can slice it at any level of detail you need for computation (that way you can merge / interact higher level structures without computing all lower level details in the process and not loosing the low details at the same time in case you'd need it later).
2) When things evolved to enough complexity to allow first consciousness to join PMR (support them), each such consciousness at the moment it joins PMR gets it's own data stream that renders the PMR "fuzzy" model into exact coherent experience of PMR (values for properties of perceived parts of PMR reality are determined as they are rendered). This rendering require measurement at level of detail determined by constrained observer's awareness - for example we don't see electromagnetic fields, so these properties are not determined when we see things unless you use some device that would translate that for your senses. This value determination constraints the level of uncertainty (fuzziness) for future measurements. So next rendering measurement could be different from previous one only within limits that rule set allows to develop. For example picture next analogy: if you look at the house and it has red roof, and you would look at it again minute later it would have most likely still red roof because under working rule set only small level of uncertainty was added over that time (decay, rain and other fuzzy things that may influenced it), but if you'd look at it after a year, it may or may not have roof of different color as red pigments could be washed off by rain and sun rays for example. It may develop uncertainty in these values (get some fuzziness again) that would be narrowed down by next measurement and everything starts again.
3) We all have our own personal PMR rendering streams, so our personal experiences of PMR (exact values computed from probability functions) may differ, even greatly at times but always within limits given by PMR rule set. Our personal rendering data stream may influence each other (narrow uncertainty to other) if they share some aspects. For example if I see the house and you see the house at the same time and we will talk about it (share experience), our experiences must be coherent (for example if rule set allows noticeable difference when each of us will get values from opposite sides of probability curve, it would create disturbing discrepancy to our experience when shared) or plausibly deniable (see Psi Uncertainty Principle and system cheating). But when our experiences doesn't overlap in any way, system doesn't care whether they are consistent to each other, they must be each consistent only to PMR rule set.