Are you remembering that with conditional pricing, you cannot interpret the main effects in the same way you do for standard conjoint? Even though I have years of experience with this, I sometimes catch myself getting ready to make this mistake when I quickly look at utilities estimated for Conditional Pricing designs.
Let me illustrate what I mean:
Let's say that you have an attribute, which is "size" that has three levels (small size, medium size, and big size). All else equal including price (main effects), you assume that bigger sizes should always be preferred to smaller sizes.
But, bigger sizes carry much higher prices in the real market, so you create a conditional pricing design that has levels of price to show with small size, other levels of price with med, and other prices with big size. You follow the design suggestions in our manual to make each price variation be a constant % down or % up from an average price.
Now, you estimate your utilities using logit, LC, or HB and you get utilities for the size attribute suggesting respondents don't like bigger sizes than smaller sizes! Your heart sinks and you wonder what you did wrong.
You need to remember that with conditional pricing designs, you have to interpret the size attribute considering the average price that the size attribute levels were shown with! The utility for "big size" may be lower than "medium size" because on average for the respondents they don't believe that the big size is worth the average price difference that is designed into the questionnaire for big relative to medium.
The market simulator understands these issues and does not lead to confusion. If you put three products in the simulator (a small size product, medium size, and large size) and then specify price levels in the simulator, then the simulator adds the part-worth utilities together and you get the proper results in the shares of preference.
And, if you wanted to see what the effect of size is holding everything (including price) constant, you would need to specify three products in the simulator, each at the same absolute price (this is only possible if your three price ranges take in a common absolute price). If, for example, price 3 for the small is the same absolute price as price 2 for medium and price 1 for big, then you could run a market simulation wherein the price was held constant for the three sizes. Only then could you see what respondents felt about size, holding everything else constant. You should then see that respondents prefer bigger sizes over smaller sizes if everything else including price was held constant.
Don't just constrain the "size" attribute in the main effects because you are getting confused regarding why you are seeing a "reversal" in your main effects! This would seriously hurt your model fit and ruin your data!