Given that the players did not register any suspicion when Zark asked for his x-bow back, should I have rolled a Bluff check for him vs PC Passive Insight? I don't normally keep a record of PC Passive scores so this would have been fairly obvious.
I would not have.
Like Passive Perception, I'm not too keen on Passive Insight. It's a great big "Hey Mr. DM, protect me from my own stupidity" tool.
There are several things that can happen during an interaction:
1) The DM can just ignore the entire Passive Insight concept. He describes the scene and the players can believe him, or not believe him, but they don't bother to go out of their way to use Active Insight.
2) The DM uses Passive Insight and one or more PCs are successful. I tend to make my first roll on this behind the screen before the conversation even starts, not mid-sentence because it's less obvious what I am doing that way. He describes the scene and the players now know for a fact that the NPC is lying or hiding something or whatever, and the scene definitely goes into the direction .
3) The DM uses Passive Insight and none of the PCs are successful. He describes the scene and the players can believe him, or not believe him, but they don't bother to go out of their way to use Active Insight. Same result as #1.
4) The players wise up and use Active Insight checks at various times in the conversation.
Here's my issue. If the DM uses Passive Insight, there is almost always one or more players with a PC with a high Insight at the table. That means that as a general rule, #2 will occur more often than #3 because most NPCs do not have a huge bluff. In your example (not sure of the level of your PCs), let's say that it is +10 vs. Zark's +8 (61.75%).
If the players do decide to use Active Insight, then the odds are even higher. Not only does the highest Insight at the table player have a 61.75% chance to be successful, the lower Insight players could roll extremely high. Maybe an 80% chance of being successful with multiple players rolling.
So when using both Passive and Active Insight within the same interaction, the odds of not being misled by the NPC are 39.25% * 20% or 7.85%. In other words, the players are going to know that the NPC is lying or hiding something or whatever over 92% of the time (85% if the DM only allows one Active Insight check from one player). It's bad enough that they are going to know this 80% of the time if they bother with their Active Insight checks at all and there's a fair chance that they will suspect it with absolutely no rolls made whatsoever.
So no. I don't think you should have used a Passive Insight check at all to protect your players from themselves. The encounter turned out interesting and the odds of the players being misled in the future as your players start asking for more Active Insight checks will drop quite a bit. Even if they do not do so, they will at least now be more suspicious of NPC motives.
Going back to the earlier discussion, this encounter of yours would have been a lot more interesting if your players would have asked for Active Insight rolls, you rolled those dice behind the screen, and you told a few players that their PCs are convinced that the dwarf is telling the truth, and you told a few different players that their PCs are convinced that the dwarf is lying.
That then leads to a lot more interesting roleplaying session as the players start grilling the dwarf or arguing amongst themselves (in character) than the players asking for Active Insight checks and the vast majority of the time, the DM states that the NPC is lying. zzzzz
By using Active Insight checks the way you prefer to do it (i.e. never give a negative result for a terrible roll), by allowing the players to see their own dice rolls, and/or by also allowing Passive Insight checks, the game boils down to chess. The players for the vast majority of encounters know exactly what is going on, so they are almost never misled. That takes a ton of fun out of the roleplaying encounter and the players almost never run into the very encounter that you just described.
The DM then has to resort to other ways to misled the players like the NPCs that they capture have limited or no knowledge and things like that. That too becomes pedestrian and boring. The game world is so much more vibrant and interesting if the PC information tools are not so powerful that the players are almost always aware of what is going on, and the DM doesn't have to go way out of his way to hide things, just to challenge the players.
By having Passive Perception and Passive Insight only relate mundane obvious information to the players and having the DM roll the PC's active checks for the more interesting information and clues, the game becomes one where the players are often not aware of some significant portion of what is going on and the surprise and mystery of the world can be exposed in cool unexpected ways instead of just letting the players know most of what is going on.
It's sometimes fun for a PC to rush across a room and walk into a trap. Not every encounter, not even every encounter day, but the DC to spot most traps should be low enough that most of the PCs have some fair chance to spot them with an Active Perception check, but the Perception Monkey cannot spot them with a Passive Perception checks because the trap is hidden and not obvious. It's boring if the Perception Monkey is allowed to spot the traps every time with a Passive check so the DM bumps up the DC for the traps so high that the rest of the PCs cannot ever spot them (in the case that the Perception Monkey is allowed to spot the interesting stuff with a Passive check).
The exact same thing applies to Passive Insight. It's boring if the info is just handed out to the Insight Monkey with a Passive check.