TL:DR: see the last subsection
The door problem.
Public buildings are required to display evacuation plans. That tells you where doors are. Obfuscating those plans would be a clear violation of safety and OSHA wouldn't be amused. Additionally, somebody built that facility. And they didn't do it alone, off the top of their head. You might argue a top secret facility wouldn't have plans just laying around, but the point is there most certainly will be plans somewhere. It's up to your intelligence services to find them.
If you have a lot of time, you might try to infiltrate someone that will map out the complex and/or procure the plans. The plans may not actually be stored on site, but instead in some sort of national archive. An alternative to placing someone would be to flip an existing employee, with or without their knowledge. This runs the risk of creating a triple agent however, which is a bad thing for you. I would limit the use of interrogation, since you'll never be able to know if they're telling the truth until you go on site and verify by yourself.
As a general advice, do not engage an operation with incomplete intelligence. You can plan contingencies for things you know. It's things you don't know that you need to worry about. You can't expect success if you go in not knowing where the doors are.
Side note regarding doors
As a general rule, your security can't rely on the attacker not knowing about your security measures. So it's pointless to hide doors because you should assume attackers will already know where the doors are. The way you prevent physical access isn't by hiding doors, it's by requiring identification (e.g. ID badge), and making sure your doors (and walls, and windows) can withstanding some force.
If you decide no doors are present on floor plans, or there's no floor plans at all, a valid strategy to move through the complex then is to consider a wall is just a door you haven't blown open yet. There's no problem a sufficient amount of force can't solve after all.
The enemies-disguised-as-friendlies problem.
Strobe lights are used IRL to identify friendlies. It's useful notably when you're way far away and people are just tiny smidges on a screen or a scope, and you need to shoot at the right people.
The holographic projectors will definitively be limited in the range of frequency they can display. As an example, if the projector works with visible light, it will be unable to reproduce an ultraviolet or infrared strobe light. Any other convenient frequency in the EM spectrum will do, or alternatively you could experiment with ultra/infrasounds. Obviously, emitting a signal means a knowledgeable enemy might be able to track it, so chose your signal wisely.
From there, the identification system is fairly simple: you have a small pack that emits your signal, you have a device that can read that signal (which could be the same device), and you have a way to tell your soldiers that information. And for that in particular, your soldiers presumably wear a helmet, which presumably has a visor. Since we're talking holographic illusion, I suppose adding augmented reality filters on a visor is a piece of cake. In the end, in practice, a friendly will pulse visibly, and an hologram will not.
Your signal emitter may get borked, which means you will no longer be identified as friendly, which is a problem if rules of engagements are "shoot first, full stop". You don't need anything complicated, a simple password will suffice. WWII soldiers during the Invasion of Normandy used "Flash/Thunder" or a clicker to identify each other beyond visual range, and that will do the job just fine.
The enemies-disguised-as-flower-pots problem, which also encompasses the above two problems.
You can deprive a system from the power it needs to operate with a well placed bomb or two. If the facility is powered by a power plant, blow up the power plant or simply cut the cable. You'll also have to sabotage the backup generators in a similar way, because there will be backup generators.
Alternatively, an EMP might fry out enough electronics to make the system inoperable, at least in the sections where you need to be, so that you know you will not encounter an hologram.
Of course, sabotage like this because extremely complicated if the relevant facilities are hardened and/or underground. It will also bring attention to the fact that the facility is under attack. So not necessarily the most viable option depending on your scenario.
Holograms have a tell
All of the above are things you should consider for any operation anyways, but this is the real killer.
Equip your soldiers with a microcamera (which really should already be standard), a microcomputer, and a helmet with an augmented reality visor. We already know about one flaw in the system: it can't fool two points of view. Have your soldiers move two by two. Each soldier's kit will receive the video feeds from their teammates, and will be able to compare the two feeds. If it detect the same object in different angles, everything is fine. If one feed shows a flower pot and another shows a man, flag that as something to kill and show it to your troopers.
If you need a tell that a single operator can detect, then you could go for a flicker. Something imperceptible to the human eye but very visible to a camera, provided it isn't sync on the same frequency than the holographic projectors. This isn't really much of a stretch if you've ever tried to film a screen.